Mechanical analyses of trabecular bone and its interaction...
Transcript of Mechanical analyses of trabecular bone and its interaction...
ACTAUNIVERSITATIS
UPSALIENSISUPPSALA
2019
Digital Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Science and Technology 1836
Mechanical analyses of trabecularbone and its interaction withimplants
DAN WU
ISSN 1651-6214ISBN 978-91-513-0715-2urn:nbn:se:uu:diva-385143
Dissertation presented at Uppsala University to be publicly examined in Ångström 8001,Lägerhyddsvägen 1, Uppsala, Wednesday, 18 September 2019 at 13:15 for the degree ofDoctor of Philosophy. The examination will be conducted in English. Faculty examiner:Professor Hanna Isaksson (Department of Biomedical Engineering, Lund University,Sweden).
AbstractWu, D. 2019. Mechanical analyses of trabecular bone and its interaction with implants.Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty of Science andTechnology 1836. 65 pp. Uppsala: Acta Universitatis Upsaliensis. ISBN 978-91-513-0715-2.
Bone substitute materials or implants are commonly used in the surgical treatment of bonefractures. However, severe complications are sometimes reported. In order to improve fracturetreatment where the interior, porous trabecular bone is involved, it is important to betterunderstand the mechanical properties of this bone and how it interacts with the substitutes/implants, and this was the aim of this thesis.
Since one of the key mechanical properties of trabecular bone, i.e. the elastic moduli at thetissue level, was not consistently reported in the literature, the results from four widely appliedmethods were first summarized and presented in a review paper.
Furthermore, to improve the analysis of the mechanical behavior of bone and its interactionwith implants, a new digital volume correlation (DVC) technique was proposed based on higher-order finite elements.
We further proposed a method to estimate the elastic modulus at the tissue level bycompression of single trabeculae within a synchrotron radiation micro-computed tomograph(SRµCT). Full-field displacements were estimated by DVC, which also provided boundaryconditions for a finite element model. The proposed method shows potential to estimatetrabecular mechanical properties at the tissue level.
Further, strains and cracks of a trabecular structure under compression till fracture werecharacterized at the single trabecular level, with DVC applied on high-resolution images fromSRµCT.
The effect of augmentation materials on the engagement of screws inserted into trabecularbone was evaluated in human femoral bone, with and without real-time SRµCT. A newlydeveloped tissue adhesive indicated a potential benefit of this material to the primary implantstability compared to a cement and no augmentation.
Finally, a trabecular structure of PLA/HA composite material was printed using a fuseddeposition modelling method as a preliminary step towards better synthetic models of trabecularbone. The synthetic trabecular structure was evaluated using micro-CT, compression and screwpull-out testing.
In conclusion, methods to estimate strains and mechanical properties of trabecular bone wereproposed, insights into interactions between trabecular bone and augmentation/implants weregained, as well as a first step towards a synthetic trabecular model, which may contribute tofurther mechanical analyses and/or improved clinical treatments of trabecular bone.
Keywords: Trabecular bone, Elastic modulus, Digital volume correlation, Pullout resistance,Micro-computed tomography.
Dan Wu, Department of Engineering Sciences, Applied Materials Sciences, Box 534, UppsalaUniversity, SE-75121 Uppsala, Sweden.
© Dan Wu 2019
ISSN 1651-6214ISBN 978-91-513-0715-2urn:nbn:se:uu:diva-385143 (http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-385143)
To my family
List of Papers
This thesis is based on the following papers, which are referred to in the textby their Roman numerals.
I Wu, D., Isaksson, P., Ferguson, S.J., Persson, C. (2018).Young’s modulus of trabecular bone at the tissue level: A re-view. Acta Biomaterialia, 78:1-12
II van Dijk, N.P., Wu, D., Persson, C., Isaksson, P. (2019). Aglobal digital volume correlation algorithm based on higher-order finite elements: Implementation and evaluation. Interna-tional Journal of Solids and Structures, 168:211-227
III Wu, D., van Dijk, N.P., Joffre, T., Öhman Mägi, C., Ferguson,S.J., Persson, C., Isaksson, P. (2019). A combined experimentaland numerical method to estimate the elastic modulus of singletrabeculae. (Submitted)
IV Wu, D., Isaksson, P., Persson, C. (2019). Quantification ofstrains and cracks in trabecular bone by digital volume correla-tion: A case study. (Submitted)
V Wu, D., Pujari-Palmer, M., Bojan, A., Palmquist, A., ÖhmanMägi, C., Ferguson, S.J., Isaksson, P., Persson, C. (2019). Theeffect of two augmentation materials on screw pullout re-sistance from human trabecular bone. (Submitted)
VI Wu, D., Spanou, A., Diez-Escudero, A., Persson, C. (2019).3D-printed PLA/HA composite structures as synthetic trabecu-lar bone: a feasibility study using Fused Deposition Modelling.(Submitted)
Reprints of Paper I were made with permission from the respective publish-ers. Paper II is published with open access.
Author’s contribution
I Major part of planning, data analyses and writing.
II Part of planning, evaluation and analyses.
III Major part of planning, experimental work, analyses and writ-ing.
IV Major part of planning, experimental work, analyses and writ-ing.
V Major part of planning, experimental work, analyses and writ-ing.
VI Part of planning, experimental work and major part of writing.
Contents
1. Introduction ............................................................................................... 11
2. Background ............................................................................................... 132.1 Bone ................................................................................................... 132.2 Trabecular bone .................................................................................. 162.3 Screw implants ................................................................................... 182.4 Imaging and mechanical testing ......................................................... 19
3. Summary of aims and objectives .............................................................. 22
4. Methods .................................................................................................... 234.1 Mechanical testing .............................................................................. 234.2 Micro-computed tomography ............................................................. 244.3 In-situ mechanical testing within SRμCT .......................................... 264.4 Digital volume correlation.................................................................. 27
5. Results and discussion .............................................................................. 30
Paper I .......................................................................................................... 31
Paper II ......................................................................................................... 34
Paper III ....................................................................................................... 38
Paper IV ....................................................................................................... 42
Paper V ........................................................................................................ 44
Paper VI ....................................................................................................... 48
6. Summary and conclusions ........................................................................ 49
7. Future perspectives ................................................................................... 51
Svensk sammanfattning ................................................................................ 52
Acknowledgements ....................................................................................... 54
References ..................................................................................................... 56
Abbreviations
BV/TV
CPCDVCFDMFEFFTHAHR-pQCT
Micro-CT
Bone volume fraction (bone volume /total volume)Calcium phosphate cementDigital volume correlationFused deposition modellingFinite elementFast-Fourier transformationHydroxyapatiteHigh resolution peripheral quantita-tive computed tomographyMicro-computed tomography
PLAPMMASRµCT
Tb.SpTb.Th
Polylactic acidPolymethylmethacrylateSynchrotron radiation based X-raymicro-computed tomographyTrabecular spacingTrabecular thickness
10
11
1. Introduction
Bone fractures are common in daily life. The incidence of new bone failuresis reported to be 2.68 million per year in six European countries [1] (France,Germany, Italy, Spain, Sweden and the UK). Trauma due to, for instance,falls and accidents, and bone disease are two common causes for bone frac-tures. The prevalence of osteoporosis, causing a majority of fractures, isreported to be 7 to 19% for men and 23 to 35% for women over 50 years old[1, 2], and the number of individuals exceeding the fracture threshold is ex-pected to be more than 300 million worldwide by 2040 [3]. With the agingpopulation and increasing life span, fragility fractures are becoming a barrierto mobility, independence and life quality, especially for the elderly.
In the surgical treatment of bone fractures, bone substitutes or implantsare utilized for various reasons. For instance, to treat bone fractures causedby high-impact trauma, internal implants such as plates and screws arecommonly used to hold together the bone fragments so that the bone canheal by itself. In the treatment of vertebral compression fractures, bone ce-ments, typically acrylic or calcium phosphate cements, have been injected tothe vertebral body for stabilization and pain relief [4], i.e. vertebroplasty.
However, complications may arise, for various reasons. In one of thetreatments aforementioned, for example, a substantial number of patientsreceiving vertebroplasty caused by osteoporosis develop new fractures, es-pecially in vertebrae adjacent to the treated level (Fig. 1) [5, 6]. One possiblereason is that the injected cement alters the rigidity of the vertebra and loadsto the adjacent vertebrae [7, 8]. Although numerous efforts have been put indeveloping better bone substituent materials, it is of the same significance tounderstand the mechanical behavior of bone itself, especially when the re-ported properties vary substantially [9, 10]. Screw implant loosening can beanother frequent complication, as in cases of humeral heads (Fig. 1) [11],osteoporotic vertebrae [12] and skeletal tumors [13], resulting in revision, ora second operation. One important reason for loosening is the inadequateinitial fixation of the screw, which leads to micro-motion of the implants.
12
Figure 1. (Left) Scan taken immediately after cement augmentation into multi-levelvertebrae. (Middle) Failure of the adjacent vertebra after approximately 2 weeks.Reprinted with permission from Baroud et al.[14]. (Right) Loosening of pediclescrew indicated by the clear zone. Reprinted with permission from El Saman etal.[15].
Mechanical analyses of bone and its interaction with implants are there-fore fundamental towards improvements in the treatment of bone fractures.A powerful tool in the mechanical analyses of bone [16], bone-cement [17],or bone-implant systems [18, 19], is finite element modeling, especially fortrabecular bone where the naturally complex structures complicate the anal-yses. Additionally, finite element models consume fewer bone samples com-paring to experimental tests. In a future perspective, finite element modelingmay be increasingly used in the analyses of bone through statistical modelsand patient-specific fracture treatment, from lab to clinical applications [20,21]. Crucial inputs for the models to be able to predict physical phenomenaof bone with accuracy are the bone’s mechanical properties at the tissue lev-el.
In this thesis, several aspects in the mechanics of trabecular bone and bone-implant interactions were investigated, with a focus on the elastic modulus of the trabecular bone at tissue level, strain concentrations in the trabecular bone, pullout resistance and failure mechanisms of screws from the trabecular bone with and without augmentation, and 3D-printing of syn-thetic trabecular bone. An algorithm with high accuracy was also proposed to quantitatively estimate the internal displace-ment and strain fields of the tested specimens in the mechanical tests.
13
2. Background
2.1 BoneBones, forming the skeletal system, have multiple functions for human be-ings. They support the weight of the body, protect vital organs, and allowour locomotion. Bones work as storage for minerals, and bone marrow with-in the bone continuously supply blood cells [22]. Generally, a bone is com-posed of a hard and dense outer layer, i.e. cortical bone, and more porousinterior, which is called trabecular bone or cancellous bone (Fig. 2). Thecombination of solid cortical and porous trabecular bone provides mechani-cal strength with relatively low density.
Figure 2.Illustration of cancellous and cortical bone with a longitudinal section of the femur. Reprinted with permission from Cowin [23].
In the material perspective, bone tissue consists of 70 wt% inorganic mat-ter, 20 wt% organic and 10 wt% water [24]. The inorganic phase is an im-pure calcium phosphate salt in the form of a modified hydroxyapatite(Ca5(PO4)3OH). About 90 wt% of the organic phase is a type I collagen. It iscommonly deemed that the minerals mainly contribute to the elastic proper-ties and the collagen to the non-linear properties. Any pathology or treatmentthat modifies the three components or their ratio might affect the mechanicalproperties of the bone. Cortical bone exposed to high doses of X-rays thatdestroy the collagen structure showed an adverse effect on strength and frac-ture toughness [25]. Dehydrated bone is reported to be stiffer and more brit-tle [26, 27].
Bone is a hierarchical composite material from the nanoscale to the mac-roscale (Fig. 3) [28, 29]. At the nanoscale, the minerals aggregate and form afractal-like organization [28]. The mineralized collagen fibrils assemble into
14
ordered arrays with a diameter from less than one micron to several microns.The arrays then form bundles with different patterns. Sheets of bundles stackin a ply-wood manner, but with the angular offset between adjacent bundlesheets being 40~80°, thus forming lamellar bone. The lamellar bone is foundin both trabecular bone with a motif of lamellar packets and cortical bone inthe form of osteons. A lamellar packet is an assembly of lamellae withslightly different orientations. In an osteon, the lamellae are aligned intolarge concentric rings, similar to that of wood rings. It should be noted thatdisordered fibril arrays and disordered phases also present in the hierarchicallevels of bone.
15
Figure 3. Hierarchical levels of bone with representative images, specifically: levelI: Components, II: Structural components, III: Arrays, IV: Array patterns, V: Super-structure, VI: Material patterns, VII: Tissue elements, VIII: Tissue and IX: Organ.The schematic is colored with green representing ordered materials, blue for thedisordered material. The borders of the images are color coded in the same way.Additionally, a graded color represents the presence of both materials. Reprintedwith permission from Reznikov et al. [29].
16
2.2 Trabecular boneAt the highest hierarchical level (Fig. 3), also referred to as the apparent level in this thesis, trabecular bone appears as a porous, cellulous structure, containing a network of trabeculae in the shape of small rods and plates (Fig. 4). Typically, the individual trabeculae have diameters between 100 to 300 μm, with a spacing between trabeculae of around 500 to 1500 μm [24]. The term ‘architecture’ or ‘morphology’ is commonly used to refer to the spatial topology or shape of the trabeculae. Trabecular morphology is continuously changing in vivo due to bone remodeling, a process where ma-ture bone matrix is removed and new one is formed. Bone remodeling re-sponds to mechanical loading as expressed by Wolff’s law: if the loading in the trabecular bone increased, the bone morphology will be remodeled to resist the load. The rate of remodeling is also affected by factors such as age and pathology. As a result, the trabecular morphology varies among individ-uals, anatomic sites, age groups, health states etc.
Figure 4. 3D renderings of trabecular bone from human femurs with (left) relativelylow and (right) high bone volume fraction, showing the rod and plate like trabecu-lae.
Due to the close relation between morphology and mechanical behavior incellulous materials, the trabecular morphology needs to be characterizedbefore further analyses of the mechanical properties. Widely used parame-ters include bone volume fraction (BV/TV, bone volume over total volume),trabecular thickness, trabecular spacing, surface-to-volume ratio, connectivi-ty, degree of anisotropy etc. With the application of imaging techniquesproviding 3-dimensional bone structures, for instance, micro-computed to-mography (micro-CT), these morphometric parameters are then calculatedfrom the volumetric digital images. Detailed definitions and calculations ofsome of these parameters are discussed in section 4.2
Many studies have reported correlations between morphometric parame-ters and apparent-level mechanical properties of trabecular bone, such as,stiffness [30], ultimate compression strength [31], and yield strength [32].The mechanical properties are reported to be mainly dependent on theBV/TV, which could explain about 70~89% of variations of the mechanical
17
properties. Furthermore, BV/TV together with the fabric tensor, i.e. a secondrank tensor that characterizes the anisotropic architecture of the trabecularbone, have been reported to improve the prediction accuracy to 90~98% [30-32].
It should be noted that the apparent-level mechanical behavior of the tra-becular bone depends not only on the aforementioned morphology, but also on the mechanical properties of the composing individual trabeculae, which are referred to as tissue-level properties. Mechanical tests have been per-formed at different levels to estimate the tissue-level mechanical properties of trabeculae (Fig. 5), from nanoindentation, directly on individual trabec-ulae, to a combination of experiments and finite element simulations on tra-becular structures (5~10 mm). Due to limitations of each method as well as the natural variation between subjects, the reported mechanical properties are in a wide range (Table 1), indicating a demand for further studies to pin-point these properties.
Figure 5. Methods to estimate tissue-level mechanical properties of trabecular bone.Reprinted with permission from Paper I [33].
Table 1. Mechanical properties of trabecular bone at the macroscopic and at thetissue level.
Ultimate compressivestrength [MPa]
Yield strain[%]
Elastic modulus[GPa]
Trabecularstructure 1-30 [34, 35] 0.6-1.2 [34, 36] 0.01-1 [34, 37]
Singletrabeculae 167-208 [38] 0.7-1.4 [38] 1.4-16 [38-40]
18
2.3 Screw implantsTwo types of screw implants are widely used in clinics: cortical and cancel-lous screws, as shown in Fig. 6. Cortical screws are mostly used in the di-aphysis of long bones containing mainly cortical bone. The threads of corti-cal screws are shallow and densely spaced to improve the engagement withthe cortical walls. Cancellous screws are used for epiphyses containing sub-stantial amounts of trabecular bone, and the threads are deeper and spacedmore widely to suit the cellulous structure of the trabecular bone.
The thread pitch and depth* of the commonly used cancellous screws (Ta-ble 2) are at the same scale as the trabecular thickness and spacing, whichmeans that trabecular bone around the screws cannot be considered as a con-tinuum [41], which has also been confirmed via finite element analyses [42].Studies have shown that the morphology of the trabecular bone closest to thescrew, especially BV/TV, plays a key role in the pullout stiffness of thescrew engagement [18].
Figure 6. Orthopaedic screws used in the fracture treatment at (A) femur, (B) humerus, and (C) diaphysis of the humerus. Reprinted with permission from Zdero et al. [43].
Table 2. Dimensions of some typical orthopaedic screws corresponding to the ISOstandard 5835:1991.
Outer diameter[mm]
Core diameter[mm]
Pitch[mm]
Trabecular screw 6.5 3.0 2.754.0 1.9 1.75
Cortical screw 4.5 3.0 1.753.5 2.4 1.25
* Thread pitch: the distance between the crests of two adjacent threads along the length of thescrew; Thread depth: the distance between the crest and the base of a thread along the radialdirection.
In cases where screw implants are inserted into low quality bone, bonecement is sometimes injected to improve the holding strength of the screws,as shown in Fig. 7. So far, the most studied biomaterials for augmentationare polymethylmethacrylate (PMMA) and calcium phosphate cements(CPCs) [44-56]. The augmentation with PMMA generally improves thepullout strength, for instance, from 1.8 to 2.8-fold in vertebrae using pediclescrews [49, 54-56]. However, the improvement of the augmentation withCPCs varies, from apparent reduction under certain conditions [57] to lessimprovement than PMMA [47, 49]. Considering the potential of CPCs interms of biodegradation and osseointegration [45, 58], effort has been putinto improving the understanding of the mechanisms that affect the engage-ment of the bone-screw-cement system [17, 49, 59], aiming at solutions formore consistent improvement in mechanical engagement.
Figure 7. Radiograph of a patient receiving plate and screws in the treatment oftibia fracture. Calcium phosphate cement is used around the long trabecular screws.Reprinted with permission from Larsson and Bauer [45].
With a recently reported phosphoserine-modified degradable CPC im-proving the adhesion between tissue 40-fold in wet condition compared tocommercial cyanoacrylates glue [60], the contribution of adhesive bondingbetween the bone-CPC and screw-CPC interfaces to the initial engagementof screw implant might lead to a promising augmentation material for screwimplants. However, the improvement of adhesion to the screw engagementneeds to be confirmed and the failure mechanism demands further study.
2.4 Imaging and mechanical testingThe combination of imaging with mechanical testing has been widely ap-plied to capture the trabecular bone behavior and to assist mechanical anal-yses of trabecular bone at different levels and at various resolutions (Table3). For instance, the location of trabecular fracture has been found to behighly correlated to the bone whitening under natural light in tests with both
19
20
single trabeculae and trabecular structures [61, 62]. Later, imaging tech-niques such as micro-CT have captured the internal deformation in the tra-becular bone during the mechanical testing, providing more information thanonly the surface changes as recorded by cameras or microscopes. Spatialdevelopment of the internal cracks induced during the trabecular compres-sion is thus possible to reveal [63].
Further digital image processing of the acquired images enables quantita-tive analyses of the trabecular bone. Features such as cracks can be segment-ed and calculated in volume [63]. Techniques such as point tracking or digi-tal image correlation have been adopted to estimate the displacement anddeformation on the trabecular surface without the adverse contribution of theend artifacts [26]. An extension of the image correlation to 3D volume corre-lation allows quantification of volumetric deformations in the trabecularbone.
The accuracy of the image processing methods depends on both the imagequality and the accuracy of the algorithms. The low accessibility to high-resolution images during mechanical loading and digital volume correlation(DVC) algorithms of relatively low accuracy have limited the quantitativeanalyses of the mechanics in trabecular bone undergoing physical defor-mation. Recently, images of high quality and resolution by synchrotron radi-ation based X-ray micro-CT (SRμCT) have become more readily available.A DVC strategy that estimates the global displacement field based on finiteelements has further shown improved performance to the previous localstrategy (see section 4.4 for more details) [64, 65]. The use of higher-orderfinite elements might better capture the continuous deformation in the tra-becular bone. The combination may therefore contribute to the quantificationof deformation with higher spatial resolution, for instance, at the single tra-becular level. Further analyses, such as comparison between strains fromfinite element modeling and DVC can therefore be possible.
21
Table 3. Imaging techniques that have been applied during mechanical testing formechanical analyses of trabecular bone.
Single trabeculae Trabecular structureHigh-speedcamera
Bone whitening during frac-ture [61]
Bone whitening during fracture [62]
Effect of dehydration on me-chanical properties [26]
Microscope Displacement estimation [66] Fracture of bone-cement composite[67]
HR-pQCT - 3D observation of bone failure [68]Micro-CT - Compression failure of bone [69]
Screw insertion process [70]Bone failure during screw pullout[71]
22
3. Summary of aims and objectives
The aim of this thesis was to further the understanding of trabecular bonebehavior at the tissue and apparent levels as well as in combination withimplants. Micro-CT was used to image the internal structure and the me-chanical behavior of the investigated trabecular bone and implants duringloading, providing the basis for the five experimental papers in this thesis.The specific objectives of these five papers are presented below, togetherwith that of the first, review paper:
(1) to review the literature on the elastic modulus of trabecular bone at thetissue-level and to analyze the effect of several factors on this property (Pa-per I);
(2) to propose a global DVC code based on higher-order finite elementsto estimate internal displacement fields and strains from volumetric imagesof inhomogeneous materials, such as trabecular bone (Paper II);
(3) to propose a method to estimate the elastic modulus of single trabecu-lae based on a combination of imaging during micromechanical testing,global DVC, and finite element analyses (Paper III);
(4) to quantify the strain fields and cracks at the single trabecular levelduring compression of trabecular structure based on in-situ testing and glob-al DVC (Paper IV);
(5) to investigate the effect of two CPC-based augmentation materialswith and without bonding strength, on the pullout resistance of screw im-plants from trabecular bone, in terms of apparent pullout load and crackingmechanisms by SRμCT (Paper V);
(6) to evaluate the feasibility of 3D-printing trabecular structures using Fused Deposition Modelling (FDM) and PLA/HA composites as synthetic trabecular model material (Paper VI).
23
4. Methods
4.1 Mechanical testingCompression testingCompression testing is a common experimental method to evaluate the me-chanical properties of various materials, for instance, elastic modulus E,yield strength v, or ultimate compressive strength u. The experiment isperformed by applying a continuous and increasing load to a regular shapedspecimen. The elastic modulus is calculated by E= / , representing the slopeof the linear region of the stress and strain curve. The ultimate compres-sive strength is calculated as u=F/A, where F is the maximum load untilfailure and A is the cross-sectional area.
Three compression tests were performed in this thesis: (1) continuouscompression of cylindrical bulk materials until failure to evaluate the ulti-mate compressive strength (Paper V, VI) at the material level, following theISO 5833 standard for evaluation of acrylic bone cement [72]; (2) compres-sion of single trabeculae as a step towards the estimation of the tissue-levelelastic modulus (Paper III); (3) compression of trabecular structures to eval-uate cracks in trabecular bone (Paper IV) and apparent-level properties ofsynthetic trabecular models (Paper VI). The two tests in Paper III and IV arefurther discussed in section 4.3, as the loading was applied step-wisely, andmicro-CT was incorporated.
Pullout testingThe engagement performance of screw implants in trabecular bone or bonemodels is usually evaluated by pullout testing, where the implants are pulledoutwards from the bone according to a standard method described in ASTMF543 [73]. Due to limitations in specimen dimensions and in hardware suchas motor speed, modified pullout tests were performed in Paper V and VI.The pullout load F is defined as the peak load obtained during the testing,the pullout stiffness is the slope of the linear region of the load and dis-placement curve, and the fracture energy is the area under the load and dis-placement curve until the pullout load F has been reached.
24
4
he specimen illuminated by xreconstructed into crosse Fourier
used for further image processing, visualization or calc
Figure
XTwo types of microsources:microfromphoton fluxalmost parallel, and ofbeam hardening and cone beam artifacts can be avoidacquire high resolution images within only a few minutes.monochromatictr
24
4.2 Micro
he specimen illuminated by xreconstructed into crosse Fourier
used for further image processing, visualization or calc
Figure
X-ray sourcesTwo types of microsources:microfromphoton fluxalmost parallel, and ofbeam hardening and cone beam artifacts can be avoidacquire high resolution images within only a few minutes.monochromatictrast and reduction in
.2 Micro
he specimen illuminated by xreconstructed into crosse Fourier
used for further image processing, visualization or calc
Figure 8.
ray sourcesTwo types of microsources:micro-CTfrom the tubephoton fluxalmost parallel, and ofbeam hardening and cone beam artifacts can be avoidacquire high resolution images within only a few minutes.monochromatic
ast and reduction in
.2 Micro
he specimen illuminated by xreconstructed into crosse Fourier
used for further image processing, visualization or calc
. Illustration
ray sourcesTwo types of microsources: a tube
CT systemthe tube
photon flux.almost parallel, and ofbeam hardening and cone beam artifacts can be avoidacquire high resolution images within only a few minutes.monochromatic
ast and reduction in
.2 Micro-computed tomography
ent, the resolution of this technique
he specimen illuminated by xreconstructed into crosse Fourier
used for further image processing, visualization or calc
Illustration
ray sourcesTwo types of micro
tube-system
the tube is of cone shape,. However,
almost parallel, and ofbeam hardening and cone beam artifacts can be avoidacquire high resolution images within only a few minutes.monochromatic
ast and reduction in
computed tomography
ent, the resolution of this technique
nated by xreconstructed into crosse Fourier
used for further image processing, visualization or calc
Illustration
Two types of micro-based micro
system (Figis of cone shape,
However,almost parallel, and ofbeam hardening and cone beam artifacts can be avoidacquire high resolution images within only a few minutes.monochromatic beam and the high photon flux
ast and reduction in
computed tomography
ent, the resolution of this technique
nated by xreconstructed into crosse Fourier
processing, visualization or calc
Illustration of the process of micro
Two types of micro-CT systembased micro
Fig.is of cone shape,
However,almost parallel, and ofbeam hardening and cone beam artifacts can be avoidacquire high resolution images within only a few minutes.
beam and the high photon fluxast and reduction in noise
computed tomography
ent, the resolution of this technique
projection data of t nated by xoss
structed images are processing, visualization or calc
the process of micro
CT systembased micro
. 9 rightis of cone shape,
However, the synchrotronalmost parallel, and of muchbeam hardening and cone beam artifacts can be avoidacquire high resolution images within only a few minutes.
beam and the high photon fluxnoise
computed tomography
aphy system was devel
ent, the resolution of this technique
projection data of t nated by xrotation angles, and then ossic algorithms, for instancstructed images are processing, visualization or calc
the process of micro
CT systembased micro-C
rightis of cone shape,
the synchrotronmuch high
beam hardening and cone beam artifacts can be avoidacquire high resolution images within only a few minutes.
beam and the high photon fluxnoise of the acquired images
computed tomography
re of an object based on itsaphy system was devel
ent, the resolution of this technique
projection data of t nated by xrotation angles, and then ossic algorithms, for instancstructed images are processing, visualization or calc
the process of micro
CT systems were used in this thesis based on the XCT system
right, also referred to as SRμCTis of cone shape, continuous
the synchrotronhigher
beam hardening and cone beam artifacts can be avoidacquire high resolution images within only a few minutes.
beam and the high photon fluxof the acquired images
computed tomographyTomography is an im
graphy system was devel
olution of this technique
ol for life scie
projection data of the specimen rotation angles, and then r
the process of micro
were used in this thesis based on the XT system
also referred to as SRμCTcontinuous
the synchrotroner photon flux.
beam hardening and cone beam artifacts can be avoidacquire high resolution images within only a few minutes.
beam and the high photon fluxof the acquired images
computed tomographyTomography is an imaging techniq
ystem was develpact
Over decades of devel tion of this technique
olution is not possible for clinical
projection data of the sperotation angles, an
uction. Finally, the recostructed images are used for f ocessing, visualization or calc
the process of micro-CT
were used in this thesis based on the XT system (Fig
also referred to as SRμCTcontinuous
the synchrotron-based Xphoton flux.
beam hardening and cone beam artifacts can be avoidacquire high resolution images within only a few minutes.
beam and the high photon fluxof the acquired images
computed tomographythe internal stru
Over decades of developm hnique
scie
projection data of the specimrotation angles, and then r
CT.
were used in this thesis based on the XFig.
also referred to as SRμCTcontinuous in
based Xphoton flux.
beam hardening and cone beam artifacts can be avoidacquire high resolution images within only a few minutes.
beam and the high photon fluxof the acquired images
the internal stru
Over decades of developm que
CTprojection data of the specimen illumirotation angles, and then reconstruc
tion. Finally, the recostructed images are used for further im , visualization or calc
were used in this thesis based on the X9 left
also referred to as SRμCTin spectrum
based X-ray beam isphoton flux.
beam hardening and cone beam artifacts can be avoidacquire high resolution images within only a few minutes.
beam and the high photon flux promote the increase inof the acquired images
ternal stru
computed tomogr
Over decades of developm hnique
s by speci. Finally, the reco
structed images are used for further image ation or calc
were used in this thesis based on the Xleft) and
also referred to as SRμCTspectrum
ray beam isphoton flux. Therefore,
beam hardening and cone beam artifacts can be avoidacquire high resolution images within only a few minutes.
promote the increase inof the acquired images.
ternal stru
computed tomogr
Over decades of developm hnique
s by speci
structed images are used for further image ation or calc
were used in this thesis based on the Xand
also referred to as SRμCTspectrum,
ray beam isTherefore,
beam hardening and cone beam artifacts can be avoided, and it is possible toacquire high resolution images within only a few minutes.
promote the increase in
ternal stru
computed tomogr
Over decades of developm hnique
s by speci
structed images are used for further image ation or calc
were used in this thesis based on the Xand a synchrotron
also referred to as SRμCT)., but
ray beam isTherefore,
, and it is possible toacquire high resolution images within only a few minutes.
promote the increase in
ternal stru
computed tomogr
Over decades of developm hnique
s by speciBased reconstruction
structed images are used for further image
were used in this thesis based on the Xsynchrotron. The X
relatively lowmon
Therefore, artifacts such as, and it is possible to
acquire high resolution images within only a few minutes. Moreover, thepromote the increase in
computed tomogr
Over decades of developm hnique
Based reconstructionstructed images are used for further image processing, visualiz
were used in this thesis based on the Xsynchrotron
The X-ray beamrelatively lowmonochromatic,artifacts such as
, and it is possible toMoreover, the
promote the increase in
to reconstruct the in
computed tomogr
Over decades of developm
sectional imageBased reconstruction
structed images are used for further image processing, visualiz
were used in this thesis based on the Xsynchrotron-based
ray beamrelatively low
ochromatic,artifacts such as
, and it is possible toMoreover, the
promote the increase in
to reconstruct the in
Over decades of development, the resolution of this tec
sectional imageBased reconstruction
structed images are used for further image processing, visualiz
were used in this thesis based on the X-raybased
ray beamrelatively low in
ochromatic,artifacts such as
, and it is possible toMoreover, the
promote the increase in con-
Over decades of development, the resolution of this tec
sectional image
raybased
ray beamin
ochromatic,artifacts such as
, and it is possible toMoreover, the
n-
Tomography is an imaging technique used to reconstruct the internal struc-ture of an object based on its projections in different directions. Ever sincethe first commercial X-ray computed tomography system was developed byHounsfield, this technique has shown great impact, especially in clinicaldiagnostics. Over decades of development, the resolution of this techniquehas improved to a few microns (micro-CT) and even of 100 nm [74] (nano-computed tomography), and while this resolution is not possible for clinicalapplications, it has become a powerful tool for life sciences and materialscience.
The main steps of X-ray based micro-CT are shown in Fig. 8, where 2Dprojection data of the specimen illuminated by X-ray are collected at variousrotation angles, and then reconstructed into cross-sectional images by specif-ic algorithms, for instance Fourier-Based reconstruction. Finally, the recon-structed images are used for further image processing, visualization or calcu-lation.
Figure(left)https://serc.carleton.edu/research_education/geochemsheets/techniques/CT.html
Figurezerland
Morphometric analysisBefore the morphometric analysis,the reconstructedtissue,geneous materialvoidthe
Figure(left)https://serc.carleton.edu/research_education/geochemsheets/techniques/CT.html
μCT facilities or beamtime is
μCT imaging in th
Figurezerland
Morphometric analysisBefore the morphometric analysis,the reconstructedtissue,geneous materialvoidthe following morphometric parameters
Figure 9. Illustrations of(left) tubehttps://serc.carleton.edu/research_education/geochemsheets/techniques/CT.html
μCT facilities or beamtime is
μCT imaging in th
Figure 10zerland. (
Morphometric analysisBefore the morphometric analysis,the reconstructedtissue, orgeneous materialvoids or
following morphometric parameters
. Illustrations oftube-based micro
https://serc.carleton.edu/research_education/geochemsheets/techniques/CT.html
μCT facilities or beamtime is
μCT imaging in th
ation about the beamline can be found
10. Front view of the sample stage(Fro
Morphometric analysisBefore the morphometric analysis,the reconstructed
or togeneous material
or the material, with global or local thresholds.following morphometric parameters
. Illustrations ofbased micro
https://serc.carleton.edu/research_education/geochemsheets/techniques/CT.html
μCT facilities or beamtime is
ation about the beamline can be found
Front view of the sample stageom https://www.psi.ch/en/sls/tomcat/endstations
Morphometric analysisBefore the morphometric analysis,the reconstructed
to limit the volume of calculation togeneous material
the material, with global or local thresholds.following morphometric parameters
. Illustrations ofbased micro
https://serc.carleton.edu/research_education/geochemsheets/techniques/CT.html
μCT facilities or beamtime is
All the SR
ation about the beamline can be found
Front view of the sample stagem https://www.psi.ch/en/sls/tomcat/endstations
Morphometric analysisBefore the morphometric analysis,the reconstructed images
limit the volume of calculation togeneous material. T
the material, with global or local thresholds.following morphometric parameters
. Illustrations ofbased micro-CT system
https://serc.carleton.edu/research_education/geochemsheets/techniques/CT.html
μCT facilities or beamtime is
All the SRμ
line can be found
Front view of the sample stagem https://www.psi.ch/en/sls/tomcat/endstations
Morphometric analysisBefore the morphometric analysis,
imageslimit the volume of calculation to
Thethe material, with global or local thresholds.
following morphometric parameters
. Illustrations of comCT system
https://serc.carleton.edu/research_education/geochemsheets/techniques/CT.html
μCT facilities or beamtime is
All the SRμ
line can be found
Front view of the sample stagem https://www.psi.ch/en/sls/tomcat/endstations
Morphometric analysisBefore the morphometric analysis,
images to excludelimit the volume of calculation to
imagesthe material, with global or local thresholds.
following morphometric parameters
common configurations forCT system
https://serc.carleton.edu/research_education/geochemsheets/techniques/CT.html
μCT facilities or beamtime is
All the SRμ
etailed inform line can be found
Front view of the sample stagem https://www.psi.ch/en/sls/tomcat/endstations
Before the morphometric analysis,to exclude
limit the volume of calculation toimages
the material, with global or local thresholds.following morphometric parameters
mon configurations forCT system and (right)
https://serc.carleton.edu/research_education/geochemsheets/techniques/CT.html
μCT facilities or beamtime is
All the SRμ
ple stage is shown inetailed inform line can be found
Front view of the sample stagem https://www.psi.ch/en/sls/tomcat/endstations
Before the morphometric analysis,to exclude
limit the volume of calculation toimages are
the material, with global or local thresholds.following morphometric parameters
mon configurations forand (right)
https://serc.carleton.edu/research_education/geochemsheets/techniques/CT.html
ffect of the
was pe
etailed information abo n be found
Front view of the sample stagem https://www.psi.ch/en/sls/tomcat/endstations
Before the morphometric analysis, a volume of interest (VOI) is definedto exclude
limit the volume of calculation toare segment
the material, with global or local thresholds.following morphometric parameters
mon configurations forand (right)
https://serc.carleton.edu/research_education/geochemsheets/techniques/CT.html
However, the accessibility of SR time is
t Source (hown in
Front view of the sample stage at the TOMCAT beamline, SLS, PSI, Swim https://www.psi.ch/en/sls/tomcat/endstations
a volume of interest (VOI) is definedto exclude unnecessary
limit the volume of calculation tosegment
the material, with global or local thresholds.following morphometric parameters are usually
mon configurations forand (right) synchrotron
https://serc.carleton.edu/research_education/geochemsheets/techniques/CT.html
However, the accessibility of SRμ
l and errors in the experiment effect of the
ample stage is shown inetailed information about t und
at the TOMCAT beamline, SLS, PSI, Swim https://www.psi.ch/en/sls/tomcat/endstations
a volume of interest (VOI) is definedunnecessary
limit the volume of calculation tosegmented into
the material, with global or local thresholds.are usually
mon configurations for two types ofsynchrotron
https://serc.carleton.edu/research_education/geochemsheets/techniques/CT.html
However, the accessibility of SRμ
was pe
shown inetailed information about t be found
at the TOMCAT beamline, SLS, PSI, Swim https://www.psi.ch/en/sls/tomcat/endstations
a volume of interest (VOI) is definedunnecessary
limit the volume of calculation to a specific region of the hetered into
the material, with global or local thresholds.are usually
two types ofsynchrotron
https://serc.carleton.edu/research_education/geochemsheets/techniques/CT.html
However, the accessibility of SRμ
ies
etailed information about the beam
at the TOMCAT beamline, SLS, PSI, Swim https://www.psi.ch/en/sls/tomcat/endstations
a volume of interest (VOI) is definedunnecessary background
a specific region of the hetered into binaries that contain either
the material, with global or local thresholds.are usually calculated
two types ofsynchrotron-based micro
https://serc.carleton.edu/research_education/geochemsheets/techniques/CT.html
However, the accessibility of SRμ
ies
ront view of the sametailed information about the beam
at the TOMCAT beamline, SLS, PSI, Swim https://www.psi.ch/en/sls/tomcat/endstations)
a volume of interest (VOI) is definedbackground
a specific region of the heterbinaries that contain either
the material, with global or local thresholds. For trabecular bone,calculated
two types of microbased micro
https://serc.carleton.edu/research_education/geochemsheets/techniques/CT.html
However, the accessibility of SRμ
ies
ront view of the sametailed information about the beam
at the TOMCAT beamline, SLS, PSI, Swi
a volume of interest (VOI) is definedbackground
a specific region of the heterbinaries that contain either
For trabecular bone,calculated
microbased micro
https://serc.carleton.edu/research_education/geochemsheets/techniques/CT.html
However, the accessibility of SRμ
ies
ront view of the sametailed information about the beam
at the TOMCAT beamline, SLS, PSI, Swi
a volume of interest (VOI) is definedbackground
a specific region of the heterbinaries that contain either
For trabecular bone,calculated:
micro-CTbased micro-CT
https://serc.carleton.edu/research_education/geochemsheets/techniques/CT.html
However, the accessibility of SRμ
ront view of the sam
at the TOMCAT beamline, SLS, PSI, Swi
a volume of interest (VOI) is definedor unwanted
a specific region of the heterbinaries that contain either
For trabecular bone,
CT systems:CT. (From
https://serc.carleton.edu/research_education/geochemsheets/techniques/CT.html
mechanical propert
ront view of the sam
at the TOMCAT beamline, SLS, PSI, Swi
a volume of interest (VOI) is definedor unwanted
a specific region of the heterbinaries that contain either
For trabecular bone,
25
systems:(From
https://serc.carleton.edu/research_education/geochemsheets/techniques/CT.html)
ront view of the sam
at the TOMCAT beamline, SLS, PSI, Swi
a volume of interest (VOI) is defined inor unwanted
a specific region of the hetero-binaries that contain either
For trabecular bone,
25
systems:(From
at the TOMCAT beamline, SLS, PSI, Swit-
inor unwanted
o-binaries that contain either
For trabecular bone,
However, the accessibility of SRμCT facilities or beamtime is more lim-ited than tube-based micro-CT, and allows for much lower sample numbersand trial and errors in the experimental designs. A side effect of the highphoton flux is the dose effect of X-rays on the mechanical properties of tra-becular bone, which has been studied and needs further investigation [75-77].
All the SRμCT imaging in this thesis (Papers II-VI) was performed at theTOMCAT beamline, Swiss Light Source (SLS), Paul Scherrer Institut (PSI),Switzerland. A front view of the sample stage is shown in Fig. 10 and de-tailed information about the beamline can be found in Stampanoni et al. [78].
26
mounting the meof the micro
within theon the acquired images to quantify the strain field
volume of initiated cracks during the loading
insmaller relaxation during the holding time of the step
in turn benefits the contrast andspace in a SRμCT system all
technique has been applied to study theof
μCT:
Figureof single trabeculae, (b) compression of trabecular structure and (c) pullout ofscrew from trabecular bone.
26
the meof the micro
stepwithin theon the acquired images to quantify the strain field
of initiated cracks during the loadingthe advantages of SRμCT
inrelaxation during the holding time of the step
in turn benefits the contrast andspace in a SRμCT system all
technique has been applied to study theof
the current thesis,
Figureof single trabeculae, (b) compression of trabecular structure and (c) pullout ofscrew from trabecular bone.
By the meof the micro
t the
the acquired images to quantify the strain fieldthe of initiated cracks during the loading
the advantages of SRμCT
the relaxation during the holding time of the stepturn benefits the contrast andspace in a SRμCT system all
has been applied to study theof
current thesis,SR
Figure 11of single trabeculae, (b) compression of trabecular structure and (c) pullout ofscrew from trabecular bone.
V/TV),
he ‘sphere
By the memicro
t a theacquired images to quantify the strain field
the initiated cracks during the loadingadvantages of SRμCT
tioned the relaxation during the holding time of the stepwhich benefits the contrast and
in a SRμCT system allThis has been applied to study the
In thesis,SRμ of single trabeculaebecular
11. Illustrations for the inof single trabeculae, (b) compression of trabecular structure and (c) pullout ofscrew from trabecular bone.
V/TV),
he ‘sphere
By mechamber
condimation formed on images to quantify the strain fieldthe ted cracks during the loading
of SRμCT
the during the holding time of the stepwhich in the contrast andchamber a SRμCT system allThis has been applied to study theformation disc
In the thesis,of single trabeculae
IV) and pullout of screw
Illustrations for the inof single trabeculae, (b) compression of trabecular structure and (c) pullout ofscrew from trabecular bone.
V/TV),
he ‘sphere
By mounting chamber of
condimation within formed on the images to quantify the strain fieldthe volume of cracks during the loading
Besides of SRμCT
the smaller during the holding time of the stepwhich in turn the contrast andchamber SRμCT system allThis been applied to study the
disc
In the of single trabeculae
IV) and pullout of screw
Illustrations for the inof single trabeculae, (b) compression of trabecular structure and (c) pullout ofscrew from trabecular bone.
V/TV),
he ‘sphere
By mounting the testing setup on the sample stage inside thechamber of the it is possible to do multiple scans of the specimen
condi
formed on the images to quantify the strain fieldthe volume of during the loading
Besides the of SRμCTmajor benefit from the in
the smaller during the holding time of the stepwhich in turn contrast andchamber space in CT system allThis technique applied to study the
disc
In the current single trabeculae
and pullout of screw
Illustrations for the inof single trabeculae, (b) compression of trabecular structure and (c) pullout ofscrew from trabecular bone.
V/TV),
he ‘sphere
testing setup on the sample stage inside theis possible to do multiple scans of the specimen
formed on the to quantify the strain fieldthe volume of initia during the loading
Besides the SRμCT, benefit from the in
the smaller the holding time of the stepwhich in turn andchamber space in a system allThis technique has to study the
disc
In the current trabeculae
(Paper pullout of screw
Illustrations for the inof single trabeculae, (b) compression of trabecular structure and (c) pullout ofscrew from trabecular bone.
V/TV),
testing setup on the sample stage inside thepossible to do multiple scans of the specimen
loading analyses such as
formed on the acquired to quantify the strain fieldthe volume of initiated the loading
Besides the μCT, a from the in
the smaller relaxation the holding time of the stepwhich in turn benefits andchamber space in a SRμ allThis technique has been to study the
compression of trabeculae(Paper IV) of screw
Illustrations for the inof single trabeculae, (b) compression of trabecular structure and (c) pullout of
setup on the sample stage inside theCT, it to do multiple scans of the specimen
andanalyses such as
formed on the acquired im the strain fieldthe volume of initiated the loading
Besides the advantages of , a from the in
the smaller relaxation holding time of the stepwhich in turn benefits the andchamber space in a SRμCT allThis technique has been study the
intervertebral 85]
mechanical testscompression of
(Paper IV) and of screw
Illustrations for the in-situ setups covered in this thesis: (a) compressionof single trabeculae, (b) compression of trabecular structure and (c) pullout of
oxels within
ulae is calcul
sphere that fits the
situ mechanical chanical setup on the sample stage inside theCT, it is to do multiple scans of the specimen
such asformed on the acquired images the strain field
Besides the advantages of SR, a major the in
the smaller relaxation during ing time of the stepwhich in turn benefits the chamber space in a SRμCT This technique has been applie the
testscompression of single ae
(Paper IV) and screwgmentation (Paper V)
situ setups covered in this thesis: (a) compressionof single trabeculae, (b) compression of trabecular structure and (c) pullout of
tal voxels within
trabecular thickness (Tb.Th), n the trabeulae is calculated b e. the diameter of
sphere that fits the
situ mechanical chanical on the sample stage inside theCT, it is do multiple scans of the specimen
, track theFurther such as
formed on the acquired images to the strain field
Besides the advantages of SRμCT over tube, a major the in
the smaller relaxation during the time of the stepwhich in turn benefits the contrast chamber space in a SRμCT system for inThis technique has been applied to the
[84,
situ testscompression of single
(Paper IV) and pullout gmentation mater
situ setups covered in this thesis: (a) compressionof single trabeculae, (b) compression of trabecular structure and (c) pullout of
bone volume fr
diameter of the largestsphere that fits the marrow space
ae in a unit length, d
chanical testing the sample stage inside theCT, it is possible multiple scans of the specimen
theFurther such as
formed on the acquired images to strain fieldthe volume of initiated cracks during the
over tube, a major benefit in
the smaller relaxation during the of the step
inThis technique has been applied to
situ testscompression of single III), compression of tr
(Paper IV) and pullout of
situ setups covered in this thesis: (a) compressionof single trabeculae, (b) compression of trabecular structure and (c) pullout of
volumen
ngth, d
chanical testing setup sample stage inside theCT, it is possible to scans of the specimen
to Further as
formed on the acquired images to quantify in fieldthe volume of initiated cracks during the
tube, a major benefit from
the smaller relaxation during the holding of the stepthe imaging
ows This technique has been applied to study of concrete
situ III), compression of tr
(Paper IV) and pullout of the trabecular boneillustrated as
situ setups covered in this thesis: (a) compressionof single trabeculae, (b) compression of trabecular structure and (c) pullout of
n the trabe
er of the largest
nit length, d
chanical testing setup on stage inside theCT, it is possible to do scans of the specimen
to track Further analyses
formed on the acquired images to quantify the d
imaging , a major benefit from the testing with
the smaller relaxation during the holding the stepimagingtests with large setups.
of concrete
situ compression of trtrabecular bone
as
situ setups covered in this thesis: (a) compressionof single trabeculae, (b) compression of trabecular structure and (c) pullout of
n the trabe
er of the largest
nit length, d
chanical testing setup on the stage inside theCT, it is possible to do of the specimen
Further analyses such formed on the acquired images to quantify the
imaging over with
the smaller relaxation during the holding time stepof
with large setups.concrete
shown of microstru
situ mechanical (Paper on of tr
ar bone, as
situ setups covered in this thesis: (a) compressionof single trabeculae, (b) compression of trabecular structure and (c) pullout of
n the trabe
er of the largest
nit length, d
chanical testing setup on the inside theCT, it is possible to do multiple of the specimen
formed on the acquired images to quantify the strain
μCTsitu with
the smaller relaxation during the holding time of of the
situ large setups.fracture
of microstru
(Paper on of trfrom ar bone
, as
situ setups covered in this thesis: (a) compressionof single trabeculae, (b) compression of trabecular structure and (c) pullout of
chanical testing setup on the inside theCT, it is possible to do multiple specimen
then be peto quantify
CTsitu
the smaller relaxation during the holding time of the loading,of the
situ tests setups.fracture of
microstru
within(Paper III), on of tr
from the bone,
situ setups covered in this thesis: (a) compressionof single trabeculae, (b) compression of trabecular structure and (c) pullout of
calculated as the diamet
number of trabeculae in a u
chanical testing setup on the sample e theCT, it is possible to do multiple scans
defobe pe
, quantify
situ testing μCT isloading,
situ tests setups.fracture of
influence tru
within(Paper III), of tr
from the
situ setups covered in this thesis: (a) compressionof single trabeculae, (b) compression of trabecular structure and (c) pullout of
local thickness at a point i
calculated as the diamet
number of trabeculae in a u
chanical testing setup on the sample stage theCT, it is possible to do multiple scans of
can pe, or y
based μis
situ tests with
influence of tru
(Paper III), compressi trfrom the trabecular
situ setups covered in this thesis: (a) compressionof single trabeculae, (b) compression of trabecular structure and (c) pullout of
local thickness at a point i
calculated as the diamet
number of trabeculae in a u
chanical testing setup on the sample stage CT, it is possible to do multiple scans of the
level can then
, or to y
SRwise
he situ tests with large
influence of
ormed (Paper III), compression
from the trabecular
situ setups covered in this thesis: (a) compressionof single trabeculae, (b) compression of trabecular structure and (c) pullout of
calculated as the diamet
chanical testing setup on the sample stage inside CT, it is possible to do multiple scans of the
level can then be
, or to
SRμwise
he situ tests with large
ormed
situ setups covered in this thesis: (a) compressionof single trabeculae, (b) compression of trabecular structure and (c) pullout of
· bone volume fraction (BV/TV), bone volume over total volume,computed as the number of bone voxels over the total voxels withinthe VOI [79];
· trabecular thickness (Tb.Th), local thickness at a point in the trabec-ulae is calculated by the ‘sphere-fitting’ method, i.e. the diameter ofthe largest sphere that fits the trabecula [80];
· trabecular spacing (Tb.Sp), calculated as the diameter of the largestsphere that fits the marrow space [80];
· trabecular number (Tb.N), number of trabeculae in a unit length, de-fined as 1/(Tb.Th+ Tb.Sp);
4.3 In-situ mechanical testing within SR CTBy mounting a mechanical testing setup on the sample stage inside thechamber of a micro-CT, it is possible to do multiple scans of the specimen ata step-wise loading condition, and to track the micrometer-level deformationwithin the specimen. Further analyses such as DVC can then be performedon the acquired images to quantify the strain field [81], or to quantify thevolume of initiated cracks during the loading [82].
Besides the advantages of SR CT imaging over tube-based CT as men-tioned in section 4.2, a major benefit from the in-situ testing with SR CT isthe smaller relaxation during the holding time of the step-wise loading,which in turn benefits the contrast and clarity of the imaging. The sufficientchamber space in a SR CT system allows for in-situ tests with large setups.This technique has been applied to study the fracture of concrete [83], de-formation of intervertebral disc [84, 85], and shown influence of microstruc-tures.
In the current thesis, three in-situ mechanical tests were performed withinSR CT: compression of single trabeculae (Paper III), compression of tra-becular bone (Paper IV) and pullout of screws from the trabecular bone withor without two augmentation materials (Paper V), illustrated as Fig. 11.
27
4.4 Digital volume correlationSince the first proposal of DVC by Bay et al. [86], DVC has been used formultiple applications, for instance to estimate the deformation fields in tra-becular bone, to detect microcracking in concrete [82] and to determine theboundary conditions for micromechanical simulations of cast iron [87]. DVCcompares two image stacks containing volumetric information of an objectunder different states, which are usually referred to as reference volume Sand deformed volume s (Fig. 12). The idea of DVC is to track the positionchange, i.e. displacement, of a material point from the reference volume tothe deformed volume , utilizing the unique grayscale distribution in theobject. A robust method to calculate the displacement = is to min-imize the inverse normalized cross-correlation between the two volumes[88], as
min 1 ( )( )
(1)
where is the displacement field and denotes the Euclidean norm.
Figure 12. Reference volume and deformed volume . The position vector refersto a position of an arbitrary point in the body in its initial (reference) volume andthe position vector refers to the position of the same point in the body in its de-formed configuration. The displacement of the point is then given by the differenceof the two positions. Reprinted from Paper III.
To increase the sensitivity of the cross-correlation to local differences, the reference volume is divided into a set of sub-volumes. Minimization of the cross-correlation function (Eq.1) is then performed on the sub-volumes. If the minimization process is performed separately for each sub-volume using independent parameters for displacement interpolation, the algorithm is re-ferred to as a local DVC method (Fig. 13b). In a global algorithm (Fig. 13c), the displacement field is estimated with all sub-volumes connected and in-terpolated with a shared set of parameters. The continuous displacement fields given by the global algorithms enable the direct differentiation to cal-culate the strain fields. While in the local algorithms, the displacement fields are interpolated and smoothed before calculation of strain fields, due to the potential discontinuities in the displacement fields. Studies have reported improved performance by using a global algorithm [64, 65].
28
Figurealgorithm. Reprinted from Paper II.
ution of the correlation method might be too low
nt field, which is obtained by a rigid registration process
f detail in the images.rithm to track the detail with reliability. Commo
and artifacts such as rings or streaks add non
placement fields might allev
28
Figurealgorithm. Reprinted from Paper II.
yscale variation. However, if the subution of the correlation method might be too low
gorithms, the DVC is solved with mu
nt field, which is obtained by a rigid registration process
f detail in the images.rithm to track the detail with reliability. Commo
and artifacts such as rings or streaks add non
placement fields might allev
Figure 13algorithm. Reprinted from Paper II.
yscale variation. However, if the subution of the correlation method might be too low
gorithms, the DVC is solved with mu
nt field, which is obtained by a rigid registration process
f detail in the images.rithm to track the detail with reliability. Commo
and artifacts such as rings or streaks add non
placement fields might allev
13. Illustration showing the difference between a local and a global DVCalgorithm. Reprinted from Paper II.
yscale variation. However, if the subution of the correlation method might be too low
gorithms, the DVC is solved with mu
nt field, which is obtained by a rigid registration process
f detail in the images.rithm to track the detail with reliability. Commo
and artifacts such as rings or streaks add non
placement fields might allev
. Illustration showing the difference between a local and a global DVCalgorithm. Reprinted from Paper II.
correlation
yscale variation. However, if the subution of the correlation method might be too low
gorithms, the DVC is solved with mu
correlation
nt field, which is obtained by a rigid registration processs
f detail in the images.rithm to track the detail with reliability. Commo
and artifacts such as rings or streaks add non
placement fields might allev
. Illustration showing the difference between a local and a global DVCalgorithm. Reprinted from Paper II.
yscale variation. However, if the subution of the correlation method might be too low
gorithms, the DVC is solved with mu
ained by a rigid registration processs
f detail in the images.rithm to track the detail with reliability. Commo
and artifacts such as rings or streaks add non
placement fields might allev
. Illustration showing the difference between a local and a global DVCalgorithm. Reprinted from Paper II.
yscale variation. However, if the subon method might be too low
gorithms, the DVC is solved with mu
ained by a rigid registration processs
f detail in the images.rithm to track the detail with reliability. Commo
and artifacts such as rings or streaks add non
placement fields might allev
. Illustration showing the difference between a local and a global DVCalgorithm. Reprinted from Paper II.
, if the subon method might be too low
gorithms, the DVC is solved with mu
valued displaceme ained by a rigid registration processs
he algorithms but also on the
lity. Commo
ings or streaks add non
o a non
might allev
. Illustration showing the difference between a local and a global DVCalgorithm. Reprinted from Paper II.
due to the lack of gra , if the subtoo large, spatial resol on method might be too low
gorithms, the DVC is solved with mu
ng point is provided by an integervalued displaceme ained by a rigid registration process
he algorithms but also on the
it is difficult for the algo lity. Commo
ings or streaks add non
polation of grayo a non
of smoothing in the dis might allev
. Illustration showing the difference between a local and a global DVCalgorithm. Reprinted from Paper II.
due to the lack of gra , if the subtoo large, spatial resol on method might be too low
gorithms, the DVC is solved with mu
ng point is provided by an integervalued displaceme ained by a rigid registration process
volume using Fahe algorithms but also on the
image quality or amount oit is difficult for the algo lity. Commo
analyses. System noises ings or streaks add non
polation of grayo a non
of smoothing in the dis might allev
. Illustration showing the difference between a local and a global DVC
n of the dis
too large, spatial resol ethod might be too lowher analysis. Ther thms, the DVC is solved with mu
tiple iterations and
lem, inimum rather than global min
ing point is provided by an integerregistration process
n the algorithms but also on the
it is difficult for the algori detail with reliability. Commo
analyses. System noises an or streaks add nonphysical detail to the ima y
he interpolation of grayd to a non
of smoothing in the displacem allev
ithout physical
. Illustration showing the difference between a local and a global DVC
ompromise between crossn of the dis
too large, spatial r might be too lowhms, the DVC is solved with mu
the risk of reac m rather than global mind by an integer
valued displacement field, which i gistration process
n the algorithms but also on theimage quality or am es.
to a non
of smoothing in the displace ight allev
hout physical
. Illustration showing the difference between a local and a global DVC
ise between crossn of the dis
too large, spatial resolution of the cor d might be too lowher analysis. Therefore, in som
starting point
tthe risk of reaching l global min
, the star ovided by an integervalued displacement field, whic a rigid registration process
e algorithms but also on theimage quality or amount
physical detail to the ime interpolation of gray
voxel translations, rotati ns.of smoothing in the displace
tested with two
. Illustration showing the difference between a local and a global DVC
etween crossdis
t not be
too large, spatial resolution of the co d might be too low
olution of the
the risk of reaching l n global minnteger
ut also on theimage quality or amount of detail in th
liability. Commo
voxel translatio ations.of smoothing in the displaceme ev
h two
without physical
. Illustration showing the difference between a local and a global DVC
disot be
due to the lack of gratoo large, spatial resolution of the correlati oo low
her analysis. Therefore, in some al
on of the
, the starti d by an integervalued displacement field, which is obt ation process
depends not only on t so on the
it is difficult for the algo lity. Commo
analyses. System noises
polation of grayvolume is subjected t
of smoothing in the displacement fields
. Illustration showing the difference between a local and a global DVC
ot bedue to the lack of grayscale variation. However
too large, spatial resolution of the correlati oo lowher analysis. Therefore, in some al
on of the
, the starti d by an integervalued displacement field, which is obt ation process
depends not only on t so on the
it is difficult for the algo lity. Commo
analyses. System noises and artifacts such as rtherefore
volume is subjected t on, for
of smoothing in the displacement fields
. Illustration showing the difference between a local and a global DVC
resolution of the ot be
due to the lack of grayscale variation. Howevertoo large, spatial resolution of the correlati oo low
her analysis. Therefore, in some al
on of the
, the startivalued displacement field, which is obt ation process
depends not only on t so on the
it is difficult for the algo
analyses. System noises and artifacts such as r
source of error is the interon, for
. Illustration showing the difference between a local and a global DVC
ot bedue to the lack of grayscale variation. However
too large, spatial resolution of the correlati oo lowher analysis. Therefore, in some al
on of the
, the startivalued displacement field, which is obt
depends not only on t so on the
it is difficult for the algo
analyses. System noises and artifacts such as r
source of error is the interon, for
voxel translations, rotations or deformations.
. Illustration showing the difference between a local and a global DVC
too large, spatial resolution of the correlatiher analysis. Therefore, in some al
t
, the starting point is providevalued displacement field, which is obt
depends not only on t
it is difficult for the algorithm to track the detail with reliabi
analyses. System noises and artifacts such as rgenerate a
source of error is the interon, for
. Illustration showing the difference between a local and a global DVC
correlation might n
too large, spatial resolution of the correlation method might be ther analysis. Therefore, in some al
. an
, the starting point is providevalued displacement field, which is obtained by a rigid registr
depends not only on t
it is difficult for the algorithm to track the detail with reliabi
analyses. System noises and artifacts such as r
source of error is the inter
. Illustration showing the difference between a local and a global DVC
too large, spatial resolution of the correlation method might be ther analysis. Therefore, in some al
he spatial resoluti
, the starting point is providevalued displacement field, which is obtained by a rigid registr
depends not only on the algorithms but al
it is difficult for the algorithm to track the detail with reliabi
analyses. System noises and artifacts such as r
integer translati
. Illustration showing the difference between a local and a global DVC
placement field.
her analysis. Therefore, in some al
he spatial resoluti
, the starting point is providevalued displacement field, which is obtained by a rigid registr
depends not only on the algorithms but al
analyses. System noises and artifacts such as r
integer translati
However, the choice of sub-volume size is a compromise between cross-correlation performance and spatial resolution of the displacement field. Ifthe sub-volume is too small, a reliable cross-correlation might not be possi-ble due to the lack of grayscale variation. However, if the sub-volume size istoo large, spatial resolution of the correlation method might be too low forfurther analysis. Therefore, in some algorithms, the DVC is solved with mul-tiple iterations and progressively reduced sub-volume sizes [89]. The cross-correlation is started with a relatively large sub-volume size, and the esti-mated displacement field from the current iteration is used as a starting pointfor the next iteration with a smaller sub-volume size. The spatial resolutionof the displacement field is therefore incrementally and stably improved.
In the minimization problem, a good starting point decreases the compu-tation time and the risk of reaching local minimum rather than global mini-mum. In some studies [90], the starting point is provided by an integer-valued displacement field, which is obtained by a rigid registration processfor each sub-volume using Fast-Fourier Transform (FFT).
The DVC performance depends not only on the algorithms but also on theimage quality or amount of detail in the images. For regions with little detail,it is difficult for the algorithm to track the detail with reliability. Commonly,results of the sub-volumes with too low correlation are excluded from furtheranalyses. System noises and artifacts such as rings or streaks add non-physical detail to the images, which may therefore generate a non-physicaldisplacement field. Another source of error is the interpolation of grayscalevalues whenever a sub-volume is subjected to a non-integer translation, forinstance, sub-voxel translations, rotations or deformations. A proper extentof smoothing in the displacement fields might alleviate the non-physicalstrains without severely affecting the physical deformations.
The quality of the results of DVC algorithms are usually tested with twovolumetric datasets, obtained in several ways:
· Repeated scans (double scan) of one specimen without physical trans-lation;
29
· Repeated scans (double scan) of one specimen with physical transla-tion;
In the first three cases, the theoretical strain should be zero for the wholevolume; in the last two cases, the theoretical displacement and strain areknown. Therefore, the differences in displacement or strains between theDVC calculated values and the theoretical values are evaluated and deter-mined as accuracy and precision, which refer to the mean and standard devi-ation of the differences, respectively.
In this thesis, a global DVC algorithm based on higher-order finite ele-ments is proposed (Paper II) and then applied to estimate the deformation insingle trabeculae (Paper III) and trabecular bone (Paper IV) based on thehigh-resolution images from SRμCT (Figure 14).
Figure 14. Example of cross-sectional images by SRμCT from single trabecula (left,reprinted from Paper III) and trabecular structure (right, reprinted from Paper IV).
1 mm
· Generating one volumetric dataset by virtually translating the datasetwith known displacement field, usually non-integer/sub-voxel transla-tion;
· Generating one volumetric dataset by virtually deforming the datasetwith known strain field.
30
5. Results and discussion
The results of this thesis, as well as specific discussion, are presented as briefsummaries of each paper. The imaging, mechanical and analysis techniquesinvolved in the papers are listed in Table 4.
Table 4
Paper I II III IV V VI× ×
× × × × ×
× × × ×
× ×
× × ×
Micro-CT SRµCTIn-situ mechanical testingEx-situ mechanical testingDVCFinite element analysis ×
31
Paper I - Young’s modulus of the trabecularbone at the tissue level: A review
The elastic modulus of trabecular bone reported in the literature lies in awide range, covering two orders of magnitude. With an aim to narrow thisrange, and/or provide possible improvement suggestions for the test meth-ods, trabecular elastic moduli estimated by four different test methods weresummarized, namely single trabeculae tests, ultrasonic tests, nanoindentationand numerical analyses, advantages and limitations of the methods werediscussed.
Micro-mechanical tests such as tension or bending directly on the singletrabeculae are straight forward methods to estimate the mechanical proper-ties of tissue-level properties of trabecular bone. Due to the small size of thesingle trabeculae, general limitations are present in the sample preparationand handling of the trabeculae, which also have an inhomogeneous geome-try, and the very measurement of the trabecular deformation. For tensiletests, specifically, glue attaching the trabeculae to the grips enhances thedifficulties in accurately measuring only the deformation of trabeculae. Fur-thermore, misalignment of the central line of the specimen with the loadingaxis may lead to uncontrolled shear stress. Bending might be less affected bythese factors, however, the elastic modulus calculated from bending is moresensitive to the specimen geometry.
Based on a relationship between the elastic modulus and the velocity ofwave propagation in the specimen, ultrasonic tests are also applied to esti-mate trabecular elastic modulus. The larger dimensions of specimens de-crease the sensitivity to sample handling. However, the ultrasonic waves canbe strongly attenuated due to the high porosity of the trabecular bone, whichcomplicate the estimation of elastic modulus from this test.
Nanoindentation measures the elastic modulus of trabecular bone at asubmicron structural scale, and has been applied across species, anatomicsites, health states of the donors and specimen conditions. The measuredelastic modulus was reported to be affected by factors such as roughness ofthe sample surface and indentation parameters. Another feature that shouldbe pointed out is the indentation locations. Indentations, for instance, nearlocal defects which are of similar scale as the indenter such as lacunae andosteocyte, are excluded for uncontrolled contact area. This might be a reasonfor the generally higher elastic modulus reported by nanoindentation than for
32
tests on single trabeculae which contains all the defects. The level of interesttherefore should be considered while presenting/citing the trabecular proper-ties.
Combining mechanical testing on trabecular bone with finite elementsimulation that takes the trabecular geometry into account has been appliedto estimate the tissue-level elastic modulus [91]. The advantages of thismethod are the convenience in sample handling and the relative insensitivityto trabecular geometry. However, limitations present due to assumptions inthe numerical models, e.g. a homogeneous material distribution and a linearelastic mechanical behavior of the trabecular bone.
The elastic modulus reported by the reviewed literature, and conditions ofthe specimens tested were summarized into tables in Paper I, with the report-ed values ranging from 1 to 22.3 GPa, as a result of physiological and patho-logical conditions as well as scale and other methodology related limitations.Distribution of the elastic modulus was also plotted as a function of testingmethod, anatomic sites and species. However, firm conclusions with statisti-cal significance were not able to be drawn. After applying two exclusioncriteria: 1) number of specimens less than six; 2) other limitations that areconsidered to affect the reliability of the test adversely, studies reportingmore reliable data were plotted in Fig. 15, still ranging from 1.2 to 22.3 GPa.It is expected that, with continuous improvements in the resolution and accu-racy of the test methods, and more methods and studies reported, the me-chanical properties of trabecular bone and its variations among testing condi-tions, species and other factors will be better understood.
Figurethe exclusion criteriastudy remained for each species, and were hence not included for clarity.from Paper
Figurethe exclusion criteriastudy remained for each species, and were hence not included for clarity.from Paper
Figure 15the exclusion criteriastudy remained for each species, and were hence not included for clarity.from Paper
15. Trabecular elastic modulus reported by studies remaining afterthe exclusion criteriastudy remained for each species, and were hence not included for clarity.from Paper I
Trabecular elastic modulus reported by studies remaining afterthe exclusion criteriastudy remained for each species, and were hence not included for clarity.
I [33]
Trabecular elastic modulus reported by studies remaining afterthe exclusion criteriastudy remained for each species, and were hence not included for clarity.
[33].
Trabecular elastic modulus reported by studies remaining afterthe exclusion criteria. For the other species, i.e. dog, rabbit and ovine, only onestudy remained for each species, and were hence not included for clarity.
Trabecular elastic modulus reported by studies remaining afterFor the other species, i.e. dog, rabbit and ovine, only one
study remained for each species, and were hence not included for clarity.
Trabecular elastic modulus reported by studies remaining afterFor the other species, i.e. dog, rabbit and ovine, only one
study remained for each species, and were hence not included for clarity.
Trabecular elastic modulus reported by studies remaining afterFor the other species, i.e. dog, rabbit and ovine, only one
study remained for each species, and were hence not included for clarity.
Trabecular elastic modulus reported by studies remaining afterFor the other species, i.e. dog, rabbit and ovine, only one
study remained for each species, and were hence not included for clarity.
Trabecular elastic modulus reported by studies remaining afterFor the other species, i.e. dog, rabbit and ovine, only one
study remained for each species, and were hence not included for clarity.
Trabecular elastic modulus reported by studies remaining afterFor the other species, i.e. dog, rabbit and ovine, only one
study remained for each species, and were hence not included for clarity.
Trabecular elastic modulus reported by studies remaining afterFor the other species, i.e. dog, rabbit and ovine, only one
study remained for each species, and were hence not included for clarity.
Trabecular elastic modulus reported by studies remaining afterFor the other species, i.e. dog, rabbit and ovine, only one
study remained for each species, and were hence not included for clarity.
Trabecular elastic modulus reported by studies remaining afterFor the other species, i.e. dog, rabbit and ovine, only one
study remained for each species, and were hence not included for clarity.
Trabecular elastic modulus reported by studies remaining afterFor the other species, i.e. dog, rabbit and ovine, only one
study remained for each species, and were hence not included for clarity.
Trabecular elastic modulus reported by studies remaining afterFor the other species, i.e. dog, rabbit and ovine, only one
study remained for each species, and were hence not included for clarity.
Trabecular elastic modulus reported by studies remaining afterFor the other species, i.e. dog, rabbit and ovine, only one
study remained for each species, and were hence not included for clarity.
Trabecular elastic modulus reported by studies remaining after applyingFor the other species, i.e. dog, rabbit and ovine, only one
Reprinted
33
applyingFor the other species, i.e. dog, rabbit and ovine, only one
Reprinted
33
applyingFor the other species, i.e. dog, rabbit and ovine, only one
Reprinted
34
Paper II - A global digital volume correlationalgorithm based on higher-order finiteelements: Implementation and evaluation
To estimate internal deformation of trabecular bone, we proposed a globalDVC technique with higher-order finite element (27-node brick elements) tointerpolate the global displacement field, hereafter, referred to as GDVC-UU. Overlapping sub-volumes were added for more robust solutions andcurvature penalization was applied for regularization. The cross-correlationproblem became a minimization with the objective function
min [ ( )] + ( ) , (2)where and are the displacement vector for a sub-volume e and all sub-volumes, respectively, represented by the displacements at the nodes with alldegrees of freedom; ( ) is a normalized cross-correlation value of sub-volume evaluated analogously as
( )( )
( ) = (3) is the total number of sub-volumes, p is the regularization factor to adjust
the weight of curvature penalty, and the curvature penalty ( ) is definedas
( ) = , , d , (4)
where = d is the volume of sub-volume , is the displacementwhich can be represented by , and ( )/ is the spatial derivative.
The algorithm was applied to trabecular datasets undergoing various dis-placement fields in order to evaluate the code’s performance, including
· virtually (Gaussian) deformed trabecular bone,· virtually translated (in sub-voxel) datasets for interpolation error,· double scan for systematic and random errors,· physically deformed datasets for overall performance.Two trabecular datasets with relatively low and high resolution were test-
ed, referred to as Tozzi-2017 and UU-2018 (Fig. 16), respectively. The re-sults were compared with a commercial global DVC module in Avizo(Thermo Fisher Scientific Inc., Massachusetts , U.S.). Here, results from vir-tually deformed trabeculae and double scan are presented to discuss somebasic aspects of the proposed DVC algorithm.
,
Figureof TozziReprinted from Paper II
werewerethe same node spacingvolume size, while in Avizo it is the same as the subshowed that tbetter than Avizowith the 27
sented insubaccuracy, whilecurplacement field was also penalizedlocally high curvature (
Figureof TozziReprinted from Paper II
Displacements along two lines across the center of the spatial regionwerewerethe same node spacingvolume size, while in Avizo it is the same as the subshowed that tbetter than Avizowith the 27
Effects of the parameterssented insub-accuracy, whilecurvature in the displacement fieldplacement field was also penalizedlocally high curvature (
Figure 16of Tozzi-Reprinted from Paper II
Displacements along two lines across the center of the spatial regionwere deformedwere compared with the exact solutionthe same node spacingvolume size, while in Avizo it is the same as the subshowed that tbetter than Avizowith the 27
Effects of the parameterssented in
-volume sizeaccuracy, while
vature in the displacement fieldplacement field was also penalizedlocally high curvature (
16. Cross sections of the-2007 (144×144×144) and (right) a slice of UU
Reprinted from Paper II
Displacements along two lines across the center of the spatial regioneformed
compared with the exact solutionthe same node spacingvolume size, while in Avizo it is the same as the subshowed that tbetter than Avizowith the 27-
Effects of the parameterssented in Fig
volume sizeaccuracy, while
vature in the displacement fieldplacement field was also penalizedlocally high curvature (
Cross sections of the2007 (144×144×144) and (right) a slice of UU
Reprinted from Paper II
Displacements along two lines across the center of the spatial regioneformed
compared with the exact solutionthe same node spacingvolume size, while in Avizo it is the same as the subshowed that the proposedbetter than Avizo
-node element over the linear base functions used in Avizo.Effects of the parameters
Fig. 17volume size
accuracy, whilevature in the displacement field
placement field was also penalizedlocally high curvature (
Cross sections of the2007 (144×144×144) and (right) a slice of UU
Reprinted from Paper II
Displacements along two lines across the center of the spatial regioneformed with Gaussian displacements
compared with the exact solutionthe same node spacingvolume size, while in Avizo it is the same as the sub
he proposedbetter than Avizo,
node element over the linear base functions used in Avizo.Effects of the parameters
17. For the volumes without any noise or artifacts, tvolume size captured
accuracy, while large subvature in the displacement field
placement field was also penalizedlocally high curvature (
Cross sections of the2007 (144×144×144) and (right) a slice of UU
Reprinted from Paper II [92]
Displacements along two lines across the center of the spatial regionwith Gaussian displacements
compared with the exact solutionthe same node spacingvolume size, while in Avizo it is the same as the sub
he proposedshowing an advantage of the triquadratic base functions
node element over the linear base functions used in Avizo.Effects of the parameters
For the volumes without any noise or artifacts, tcaptured
large subvature in the displacement field
placement field was also penalizedlocally high curvature (
Cross sections of the2007 (144×144×144) and (right) a slice of UU
[92].
Displacements along two lines across the center of the spatial regionwith Gaussian displacements
compared with the exact solutionthe same node spacing. In GDVCvolume size, while in Avizo it is the same as the sub
he proposedshowing an advantage of the triquadratic base functions
node element over the linear base functions used in Avizo.Effects of the parameters
For the volumes without any noise or artifacts, tcaptured
large subvature in the displacement field
placement field was also penalizedlocally high curvature (Fig.
Cross sections of the trabecular datasets2007 (144×144×144) and (right) a slice of UU
.
Displacements along two lines across the center of the spatial regionwith Gaussian displacements
compared with the exact solutionIn GDVC
volume size, while in Avizo it is the same as the subhe proposed algorithm captured the Gaussian displacements
showing an advantage of the triquadratic base functionsnode element over the linear base functions used in Avizo.
Effects of the parameters ofFor the volumes without any noise or artifacts, tcaptured the Gaussian displacement field
large sub-volvature in the displacement field
placement field was also penalized. 17e
trabecular datasets2007 (144×144×144) and (right) a slice of UU
Displacements along two lines across the center of the spatial regionwith Gaussian displacements
compared with the exact solutionIn GDVC
volume size, while in Avizo it is the same as the subalgorithm captured the Gaussian displacements
showing an advantage of the triquadratic base functionsnode element over the linear base functions used in Avizo.
of GDVCFor the volumes without any noise or artifacts, t
the Gaussian displacement fieldvolume size could not
vature in the displacement fieldplacement field was also penalized
e-f).
trabecular datasets2007 (144×144×144) and (right) a slice of UU
Displacements along two lines across the center of the spatial regionwith Gaussian displacements
compared with the exact solutionIn GDVC-UU,
volume size, while in Avizo it is the same as the subalgorithm captured the Gaussian displacements
showing an advantage of the triquadratic base functionsnode element over the linear base functions used in Avizo.
GDVCFor the volumes without any noise or artifacts, t
the Gaussian displacement fieldume size could not
vature in the displacement fields (placement field was also penalized using
f).
trabecular datasets2007 (144×144×144) and (right) a slice of UU
Displacements along two lines across the center of the spatial regionwith Gaussian displacements
compared with the exact solutions andUU,
volume size, while in Avizo it is the same as the subalgorithm captured the Gaussian displacements
showing an advantage of the triquadratic base functionsnode element over the linear base functions used in Avizo.
GDVC-UUFor the volumes without any noise or artifacts, t
the Gaussian displacement fieldume size could not
(Figusing
trabecular datasets2007 (144×144×144) and (right) a slice of UU
Displacements along two lines across the center of the spatial regionwith Gaussian displacements
and the displacements fromUU, the node spacing is half the sub
volume size, while in Avizo it is the same as the subalgorithm captured the Gaussian displacements
showing an advantage of the triquadratic base functionsnode element over the linear base functions used in Avizo.
UU onFor the volumes without any noise or artifacts, t
the Gaussian displacement fieldume size could not
Fig. 17using regularization, especially in the
trabecular datasets used in this work2007 (144×144×144) and (right) a slice of UU
Displacements along two lines across the center of the spatial regionwith Gaussian displacements were
the displacements fromhe node spacing is half the sub
volume size, while in Avizo it is the same as the subalgorithm captured the Gaussian displacements
showing an advantage of the triquadratic base functionsnode element over the linear base functions used in Avizo.
on the performance are also prFor the volumes without any noise or artifacts, t
the Gaussian displacement fieldume size could not
17c-d)regularization, especially in the
used in this work2007 (144×144×144) and (right) a slice of UU
Displacements along two lines across the center of the spatial regionwere
the displacements fromhe node spacing is half the sub
volume size, while in Avizo it is the same as the sub-algorithm captured the Gaussian displacements
showing an advantage of the triquadratic base functionsnode element over the linear base functions used in Avizo.
the performance are also prFor the volumes without any noise or artifacts, t
the Gaussian displacement fieldume size could not capture
d). The curvature of the diregularization, especially in the
used in this work2007 (144×144×144) and (right) a slice of UU-2018 (200×200×200).
Displacements along two lines across the center of the spatial regionwere evaluated
the displacements fromhe node spacing is half the sub
-volume size.algorithm captured the Gaussian displacements
showing an advantage of the triquadratic base functionsnode element over the linear base functions used in Avizo.
the performance are also prFor the volumes without any noise or artifacts, t
the Gaussian displacement fieldcapture
The curvature of the diregularization, especially in the
used in this work2018 (200×200×200).
Displacements along two lines across the center of the spatial regionevaluated
the displacements fromhe node spacing is half the sub
volume size.algorithm captured the Gaussian displacements
showing an advantage of the triquadratic base functionsnode element over the linear base functions used in Avizo.
the performance are also prFor the volumes without any noise or artifacts, t
the Gaussian displacement fieldcapture
The curvature of the diregularization, especially in the
used in this work2018 (200×200×200).
Displacements along two lines across the center of the spatial regionevaluated. T
the displacements fromhe node spacing is half the sub
volume size.algorithm captured the Gaussian displacements
showing an advantage of the triquadratic base functionsnode element over the linear base functions used in Avizo.
the performance are also prFor the volumes without any noise or artifacts, t
the Gaussian displacement fieldsthe locally high
The curvature of the diregularization, especially in the
used in this work: (left) a slice2018 (200×200×200).
Displacements along two lines across the center of the spatial region. The results
the displacements fromhe node spacing is half the sub
volume size. Figalgorithm captured the Gaussian displacements
showing an advantage of the triquadratic base functionsnode element over the linear base functions used in Avizo.
the performance are also prFor the volumes without any noise or artifacts, the smallest
with highestthe locally high
The curvature of the diregularization, especially in the
: (left) a slice2018 (200×200×200).
Displacements along two lines across the center of the spatial regionhe results
the displacements from Avizohe node spacing is half the sub
Fig. 17algorithm captured the Gaussian displacements
showing an advantage of the triquadratic base functionsnode element over the linear base functions used in Avizo.
the performance are also prhe smallest
with highestthe locally high
The curvature of the diregularization, especially in the
35
: (left) a slice2018 (200×200×200).
Displacements along two lines across the center of the spatial region thathe results
Avizo athe node spacing is half the sub
17a-algorithm captured the Gaussian displacements
showing an advantage of the triquadratic base functions
the performance are also pre-he smallest
with highestthe locally high
The curvature of the dis-regularization, especially in the
35
: (left) a slice2018 (200×200×200).
thathe results
athe node spacing is half the sub-
-balgorithm captured the Gaussian displacements
showing an advantage of the triquadratic base functions
e-he smallest
with highestthe locally high
s-regularization, especially in the
36
Figure 17. A cross section of the resulting displacement fields u(x, y) in [voxels]along two lines A and B obtained by Avizo and GDVC-UU for a virtual Gaussiandisplacement field applied to the Tozzi-2017 dataset. Figures (a) and (b) comparethe different methods, (c) and (d) different sub-volume sizes and (e) and (f) differentamounts of regularization. The ’sv’ denotes the sub-volume size. Reprinted fromPaper II [92].
The two algorithms performed comparably at the same node spacing onimages from two repeated scans. It is also noted that accuracy and precisiondecreased with decreasing sub-volume size (Fig. 18), which is contrary tothe Gaussian cases. The increased freedom in the displacement field usingsmaller sub-volumes allows increased fitting to the physical local defor-
37
mation, but also increased fitting to the noise or artifacts, leading to non-physical deformation.
Regularization is needed to alleviate the non-physical deformation, andthe amount of regularization depends on the pattern and quality of the imag-es. To limit the non-physical strains to below 1‰ in the Tozzi-2017 doublescan, a regularization factor larger than 10 was needed, for the UU-2018dataset, a factor of 104 was needed due to the relatively larger void in onesub-volume.
Figure 18. The accuracy and precision of the calculated strain fields in [‰] ob-tained by Avizo and GDVC-UU for the Tozzi-2017 double scan (152×152×432) andthe UU-2018 double scan. Reprinted from Paper II [92].
Results showed that the proposed global DVC algorithm performs compa-rably to and slightly better than Avizo in most cases. Besides, the resultsshowed that to evaluate the performance of a DVC code in terms of sensi-tivity to noise and artifacts, interpolation error, etc., several tests cases areneeded, including double scan, virtually deformed and physically deformeddatasets.
38
Paper III - A combined experimental andnumerical method to estimate the elasticmodulus of single trabeculae
The previous studies in Papers I and II demonstrated i) a need to estimate thetissue-level elastic modulus of trabecular bone with higher accuracy, and ii)the possibility of estimating internal displacement of tested specimens withhigh accuracy by a global DVC method. A strategy combing compression ofsingle trabeculae, DVC and finite element (FE) analyses was therefore pro-posed to estimate the elastic modulus of single trabeculae. As illustrated byFig. 19, the strategy started from an in-situ compression test of single tra-beculae, in which the reaction forces of the trabeculae were recorded, andthe trabecular deformation was captured by SRμCT during step-wise load-ing. DVC on the SRμCT images numerically estimated the displacementfields, which then served as boundary conditions in FE models. The high-resolution FE models generated from the SRμCT images accounted for theirregular geometries of the trabeculae. The elastic moduli were finally esti-mated by comparing the reaction forces from the compression and the FEmodels at similar deformations.
Here, two main aspects of this method are discussed and the results onthree trabeculae are presented.
Figuresingle trabeculae.
ciaas discussed earlier, highly affect the performance.strategy based on virtual strain energy wasphysical
whereand ν=0.3, for simplicity.varied with the choice of parameters. The curvaturereflectfieldcrease in the curvaturereached. The regularization factor for this subat the value ofsubcy and precision in the subsenmaterial
Figuresingle trabeculae.
The first point is the choice of parameterscially subas discussed earlier, highly affect the performance.strategy based on virtual strain energy wasphysical
whereand ν=0.3, for simplicity.varied with the choice of parameters. The curvaturereflectfieldcrease in the curvaturereached. The regularization factor for this subat the value ofsub-cy and precision in the subsen.material
Figure 19single trabeculae.
The first point is the choice of parametersly sub
as discussed earlier, highly affect the performance.strategy based on virtual strain energy wasphysical
=whereand ν=0.3, for simplicity.varied with the choice of parameters. The curvaturereflectsfield concrease in the curvaturereached. The regularization factor for this subat the value of
-volume size and regularization factor, thecy and precision in the sub
. By doing this,material
19. Schematic: Estimating the elastic modulus from compression testing ofsingle trabeculae.
The first point is the choice of parametersly sub-volume size and regularization factor
as discussed earlier, highly affect the performance.strategy based on virtual strain energy wasphysical deformation, let the virtual strain energy
[=
and ν=0.3, for simplicity.varied with the choice of parameters. The curvature
the stability ofconsiderably
crease in the curvaturereached. The regularization factor for this subat the value of
volume size and regularization factor, thecy and precision in the sub
By doing this,material, as it
Schematic: Estimating the elastic modulus from compression testing ofsingle trabeculae.
The first point is the choice of parametersvolume size and regularization factor
as discussed earlier, highly affect the performance.strategy based on virtual strain energy was
deformation, let the virtual strain energy[
=and ν=0.3, for simplicity.varied with the choice of parameters. The curvature
the stability ofsiderably
crease in the curvaturereached. The regularization factor for this subat the value of
volume size and regularization factor, thecy and precision in the sub
By doing this,, as it
Schematic: Estimating the elastic modulus from compression testing ofsingle trabeculae. Reprinted
The first point is the choice of parametersvolume size and regularization factor
as discussed earlier, highly affect the performance.strategy based on virtual strain energy was
deformation, let the virtual strain energy
+/
and ν=0.3, for simplicity.varied with the choice of parameters. The curvature
the stability ofsiderably, leading to
crease in the curvaturereached. The regularization factor for this sub
resulting in this firsvolume size and regularization factor, the
cy and precision in the subBy doing this,
, as it keep
Schematic: Estimating the elastic modulus from compression testing ofReprinted
The first point is the choice of parametersvolume size and regularization factor
as discussed earlier, highly affect the performance.strategy based on virtual strain energy was
deformation, let the virtual strain energy+
, Ω is the volume, repeated indand ν=0.3, for simplicity.varied with the choice of parameters. The curvature
the stability of, leading to
crease in the curvaturereached. The regularization factor for this sub
resulting in this firsvolume size and regularization factor, the
cy and precision in the suba balanced combination
keeps a relatively low
Schematic: Estimating the elastic modulus from compression testing ofReprinted f
The first point is the choice of parametersvolume size and regularization factor
as discussed earlier, highly affect the performance.strategy based on virtual strain energy was
deformation, let the virtual strain energy
, Ω is the volume, repeated indand ν=0.3, for simplicity.varied with the choice of parameters. The curvature
the stability of (, leading to
crease in the curvature κreached. The regularization factor for this sub
resulting in this firsvolume size and regularization factor, the
cy and precision in the suba balanced combination
a relatively low
Schematic: Estimating the elastic modulus from compression testing offrom Paper III.
The first point is the choice of parametersvolume size and regularization factor
as discussed earlier, highly affect the performance.strategy based on virtual strain energy was
deformation, let the virtual strain energy
, Ω is the volume, repeated indand ν=0.3, for simplicity. The strain energyvaried with the choice of parameters. The curvature
)., leading to
indicatesreached. The regularization factor for this sub
resulting in this firsvolume size and regularization factor, the
cy and precision in the sub-voxel translation test (see sectiona balanced combination
a relatively low
Schematic: Estimating the elastic modulus from compression testing ofrom Paper III.
The first point is the choice of parametersvolume size and regularization factor
as discussed earlier, highly affect the performance.strategy based on virtual strain energy was
deformation, let the virtual strain energy]d
, Ω is the volume, repeated indThe strain energy
varied with the choice of parameters. The curvature. Generally, a high
, leading to lowindicates
reached. The regularization factor for this subresulting in this firs
volume size and regularization factor, thevoxel translation test (see section
a balanced combinationa relatively low
Schematic: Estimating the elastic modulus from compression testing ofrom Paper III.
The first point is the choice of parametersvolume size and regularization factor
as discussed earlier, highly affect the performance.strategy based on virtual strain energy was
deformation, let the virtual strain energy
, Ω is the volume, repeated indThe strain energy
varied with the choice of parameters. The curvatureGenerally, a highlow strain energy
indicates areached. The regularization factor for this sub
resulting in this first peakvolume size and regularization factor, the
voxel translation test (see sectiona balanced combination
a relatively low
Schematic: Estimating the elastic modulus from compression testing ofrom Paper III.
The first point is the choice of parametersvolume size and regularization factor
as discussed earlier, highly affect the performance.strategy based on virtual strain energy was
deformation, let the virtual strain energy
, Ω is the volume, repeated indThe strain energy
varied with the choice of parameters. The curvatureGenerally, a high
strain energya plateau in the strain fields has been
reached. The regularization factor for this subt peak
volume size and regularization factor, thevoxel translation test (see section
a balanced combinationa relatively low virtual strain energy
Schematic: Estimating the elastic modulus from compression testing of
The first point is the choice of parametersvolume size and regularization factor
as discussed earlier, highly affect the performance.strategy based on virtual strain energy was applied. For a volume undergoing
deformation, let the virtual strain energy
, Ω is the volume, repeated indThe strain energy
varied with the choice of parameters. The curvatureGenerally, a high
strain energyplateau in the strain fields has been
reached. The regularization factor for this subt peak. A
volume size and regularization factor, thevoxel translation test (see section
a balanced combination was obtainedvirtual strain energy
Schematic: Estimating the elastic modulus from compression testing of
The first point is the choice of parameters ofvolume size and regularization factor
as discussed earlier, highly affect the performance.applied. For a volume undergoing
deformation, let the virtual strain energy
, Ω is the volume, repeated indThe strain energy computed from DVC results
varied with the choice of parameters. The curvatureGenerally, a high
strain energyplateau in the strain fields has been
reached. The regularization factor for this sub-volume size was then chosenAmong several combinations of
volume size and regularization factor, the onevoxel translation test (see section
was obtainedvirtual strain energy
Schematic: Estimating the elastic modulus from compression testing of
of the DVC algorithm, esp(here denoted as
as discussed earlier, highly affect the performance.applied. For a volume undergoing
deformation, let the virtual strain energy
, Ω is the volume, repeated indcomputed from DVC results
varied with the choice of parameters. The curvatureregulate
strain energy . However,plateau in the strain fields has been
volume size was then chosenmong several combinations ofone that g
voxel translation test (see sectionwas obtained
virtual strain energy
Schematic: Estimating the elastic modulus from compression testing of
the DVC algorithm, esp(here denoted as
as discussed earlier, highly affect the performance. Here, a straightforwardapplied. For a volume undergoing
, Ω is the volume, repeated index denote summationcomputed from DVC results
varied with the choice of parameters. The curvature κ of the functionregulate. However,
plateau in the strain fields has beenvolume size was then chosen
mong several combinations ofthat g
voxel translation test (see sectionwas obtained
virtual strain energy
Schematic: Estimating the elastic modulus from compression testing of
the DVC algorithm, esp(here denoted as
ere, a straightforwardapplied. For a volume undergoing
(5)ex denote summation
computed from DVC resultsof the function
regulates. However,
plateau in the strain fields has beenvolume size was then chosen
mong several combinations ofthat gives
voxel translation test (see sectionwas obtained for the
virtual strain energy
Schematic: Estimating the elastic modulus from compression testing of
the DVC algorithm, esp(here denoted as
ere, a straightforwardapplied. For a volume undergoing
(5)ex denote summation
computed from DVC resultsof the function
the deformation. However,
plateau in the strain fields has beenvolume size was then chosen
mong several combinations ofs the best accur
voxel translation test (see section 4.4for the
virtual strain energy and
Schematic: Estimating the elastic modulus from compression testing of
the DVC algorithm, esp(here denoted as
ere, a straightforwardapplied. For a volume undergoing
ex denote summationcomputed from DVC results
of the functionthe deformation
. However, dramaticplateau in the strain fields has been
volume size was then chosenmong several combinations of
the best accur4.4) was ch
for the consideredand reasonable
Schematic: Estimating the elastic modulus from compression testing of
the DVC algorithm, esp), which,
ere, a straightforwardapplied. For a volume undergoing
ex denote summationcomputed from DVC results
of the functionthe deformation
dramaticplateau in the strain fields has been
volume size was then chosenmong several combinations of
the best accur) was chconsideredreasonable
39
Schematic: Estimating the elastic modulus from compression testing of
the DVC algorithm, espe-, which,
ere, a straightforwardapplied. For a volume undergoing
ex denote summationcomputed from DVC results
of the function (the deformation
dramatic in-plateau in the strain fields has been
volume size was then chosenmong several combinations of
the best accura-) was cho-consideredreasonable
39
Schematic: Estimating the elastic modulus from compression testing of
e-, which,
ere, a straightforwardapplied. For a volume undergoing
ex denote summationcomputed from DVC results
( )the deformation
n-plateau in the strain fields has been
volume size was then chosenmong several combinations of
a-o-
consideredreasonable
( )
40
accuracy in the estimated displacement fieldsubselected
Figureregularization factorsstandard deviations of the estimated displacement field whenbinations oftion datain respective set
ulae
orm differently
ratio between the discretized trabeculae a
lotted for each cross sectionudinal (trabecular length) direction, calculated as
ded.
40
accuracy in the estimated displacement fieldsub-selected
Figureregularization factorsstandard deviations of the estimated displacement field whenbinations oftion datain respective set
endency of the estimated elastic modulus on theulae
orm differently
ratio between the discretized trabeculae a
lotted for each cross sectionudinal (trabecular length) direction, calculated as
ded.pendency of the estimated elastic moduli on
accuracy in the estimated displacement field-volume size of 40×40×40 voxels and a regularization factor of 100 was
selected
Figure 20regularization factorsstandard deviations of the estimated displacement field whenbinations oftion datain respective set
endency of the estimated elastic modulus on theulae
orm differently
ratio between the discretized trabeculae a
lotted for each cross sectionudinal (trabecular length) direction, calculated as
ded.pendency of the estimated elastic moduli on
accuracy in the estimated displacement fieldvolume size of 40×40×40 voxels and a regularization factor of 100 was
selected for subs
20. Virtual strain energy Π and curvature κ at various subregularization factorsstandard deviations of the estimated displacement field whenbinations of regularization factortion dataset (bottom)in respective set
endency of the estimated elastic modulus on theulae
orm differently
ratio between the discretized trabeculae a
lotted for each cross sectionudinal (trabecular length) direction, calculated as
placement from FE analyses,
ded.pendency of the estimated elastic moduli on
accuracy in the estimated displacement fieldvolume size of 40×40×40 voxels and a regularization factor of 100 was
for subs
Virtual strain energy Π and curvature κ at various subregularization factorsstandard deviations of the estimated displacement field when
regularization factor(bottom)
in respective set-up in the limit
endency of the estimated elastic modulus on theulae
orm differently
ratio between the discretized trabeculae a
lotted for each cross sectionngth) direction, calculated as
placement from FE analyses,
ce
ded.pendency of the estimated elastic moduli on
accuracy in the estimated displacement fieldvolume size of 40×40×40 voxels and a regularization factor of 100 was
for subsequent analyses
Virtual strain energy Π and curvature κ at various subregularization factorsstandard deviations of the estimated displacement field when
regularization factor(bottom). The bold lines indicate where the curvature κ has its first peak
up in the limit
endency of the estimated elastic modulus on theulae
orm differently
ratio between the discretized trabeculae aference in cross
lotted for each cross sectionngth) direction, calculated as
placement from FE analyses,
ce
pendency of the estimated elastic moduli on
accuracy in the estimated displacement fieldvolume size of 40×40×40 voxels and a regularization factor of 100 was
equent analyses
Virtual strain energy Π and curvature κ at various subregularization factors αstandard deviations of the estimated displacement field when
regularization factor. The bold lines indicate where the curvature κ has its first peak
up in the limit
endency of the estimated elastic modulus on theulae
tized trabeculae aference in cross
lotted for each cross sectionalong the longit ngth) direction, calculated as
placement from FE analyses,
ce
pendency of the estimated elastic moduli on
accuracy in the estimated displacement fieldvolume size of 40×40×40 voxels and a regularization factor of 100 was
equent analyses
Virtual strain energy Π and curvature κ at various subfor the physically deformed data
standard deviations of the estimated displacement field whenregularization factor
. The bold lines indicate where the curvature κ has its first peakup in the limit
endency of the estimated elastic modulus on theulae
denoting the length tized trabeculae aference in cross
lotted for each cross sectionalong the longit ngth) direction, calculated as
placement from FE analyses,
ce
regions were exclulastic moduli on
accuracy in the estimated displacement fieldvolume size of 40×40×40 voxels and a regularization factor of 100 was
equent analyses
Virtual strain energy Π and curvature κ at various subor the physically deformed data
standard deviations of the estimated displacement field whenregularization factor
. The bold lines indicate where the curvature κ has its first peakup in the limit
d elastic modulus on theulae
enclosed with a type of epoxyglue, which might def
denoting the length tized trabeculae aference in cross
lotted for each cross sectionalong the longit ngth) direction, calculated as
placement from FE analyses,
ce
The more scattered de lastic moduli on
accuracy in the estimated displacement fieldvolume size of 40×40×40 voxels and a regularization factor of 100 was
equent analyses (Fig
Virtual strain energy Π and curvature κ at various subor the physically deformed data
standard deviations of the estimated displacement field whenregularization factors α and sub
. The bold lines indicate where the curvature κ has its first peak0
The other point is the dep d elastic modulus on the
enclosed with a type of epoxyglue, which might def
denoting the length tized trabeculae aference in cross
lotted for each cross sectionalong the longit ngth) direction, calculated as
es,
ce
The more scattered de lastic moduli on
accuracy in the estimated displacement fieldvolume size of 40×40×40 voxels and a regularization factor of 100 was
Fig.
Virtual strain energy Π and curvature κ at various subor the physically deformed data
standard deviations of the estimated displacement field whenα and sub
. The bold lines indicate where the curvature κ has its first peak0. Reprinted
The other point is the d stimated elastic modulus on thelength of the discretitip of the trabecul severely during compression. Add
enclosed with a type of epoxy
tted for each cross sectionalong the longitudinal (trabecul n, calculated as
is the longitudinal m FE analyses,tudinal displacemen ptcross section k in the two dis elds.
c. After exclud dels
accuracy in the estimated displacement fieldvolume size of 40×40×40 voxels and a regularization factor of 100 was
20)
Virtual strain energy Π and curvature κ at various subor the physically deformed data
standard deviations of the estimated displacement field whenα and sub-
. The bold lines indicate where the curvature κ has its first peakReprinted
The other point is the depe ic modulus on the
glue, which might def
denoting the length ratio between t abeculae a). The relati n cross
r each cross sectionalong the longitudinal (trabecu ection, calculated as
is the longitudi om FE analyses,
s.b shows that the rel
accuracy in the estimated displacement field.volume size of 40×40×40 voxels and a regularization factor of 100 was
).
Virtual strain energy Π and curvature κ at various subor the physically deformed data
standard deviations of the estimated displacement field when-volume size
. The bold lines indicate where the curvature κ has its first peakReprinted f
The other point is the dep ated elastic modulus on thel
the trabecul h a type of epoxy
with
is the longitudinal FE analyses,
The more scattere c moduli onc. After exclud odels
. In this study, avolume size of 40×40×40 voxels and a regularization factor of 100 was
Virtual strain energy Π and curvature κ at various subor the physically deformed data
standard deviations of the estimated displacement field whenvolume size
. The bold lines indicate where the curvature κ has its first peakfrom Paper
The other point is the de ated elastic modulus on the
tip of the trabecula sometim compression. Addthe trabecul th a type of epoxy
ends of the trabecula were gradually FE model, with
along the longitudinal (trabecular
is the longitudinal nalyses,
nd damaged tip
The more scattered dependency of the c moduli on
In this study, avolume size of 40×40×40 voxels and a regularization factor of 100 was
Virtual strain energy Π and curvature κ at various subor the physically deformed data
standard deviations of the estimated displacement field whenvolume sizes for
. The bold lines indicate where the curvature κ has its first peakrom Paper
The other point is the dep e
the trabecula was pe of epoxyboth
denoting the length ratio between the discre ae a). The relative dif
between the FE analyses and the DVC are palong the longitudinal (trabecular le ulated as
is the longitudinal dis es,
The more scattered de lastic moduli on
In this study, avolume size of 40×40×40 voxels and a regularization factor of 100 was
Virtual strain energy Π and curvature κ at various subor the physically deformed dataset
standard deviations of the estimated displacement field when applying various cos for
. The bold lines indicate where the curvature κ has its first peakrom Paper III.
The other point is the dep e
the trabecula was pe of epoxyboth
denoting the length ratio between the discre ae a). The relative dif
between the FE analyses and the DVC are palong the longitudinal (trabecular le
is the longitudinal dis es,
The more scattered de
In this study, avolume size of 40×40×40 voxels and a regularization factor of 100 was
Virtual strain energy Π and curvature κ at various subset (top)
applying various cothe sub
. The bold lines indicate where the curvature κ has its first peakIII.
The other point is the dependency of the estimate e
the trabecula was pe of epoxyboth
denoting the length ratio between the discre
between the FE analyses and the DVC are palong the longitudinal (trabecular length) direction, calc
is the longitudinal dis
The more scattered dependency of the estimated e
In this study, a combination ofvolume size of 40×40×40 voxels and a regularization factor of 100 was
Virtual strain energy Π and curvature κ at various sub-volume(top); t
applying various cothe sub-
. The bold lines indicate where the curvature κ has its first peak
The other point is the dependency of the estimate e
the trabecula was pe of epoxyboth
denoting the length ratio between the discretized trabecul
between the FE analyses and the DVC are palong the longitudinal (trabecular length) direction, calc
is the longitudinal displacement from FE analys
The more scattered dependency of the estimated e
combination ofvolume size of 40×40×40 voxels and a regularization factor of 100 was
volumethe mean and
applying various co-voxel transl
. The bold lines indicate where the curvature κ has its first peak
The other point is the dependency of the estimate e
the trabecula was both
between the FE analyses and the DVC are palong the longitudinal (trabecular length) direction, calc
The more scattered dependency of the estimated e
combination ofvolume size of 40×40×40 voxels and a regularization factor of 100 was
volume sizes andhe mean and
applying various covoxel transl
. The bold lines indicate where the curvature κ has its first peak
The other point is the dependency of the estimate e
the trabecula was enclosed with a tyboth
between the FE analyses and the DVC are p
The more scattered dependency of the estimated e
combination ofvolume size of 40×40×40 voxels and a regularization factor of 100 was
sizes andhe mean and
applying various com-voxel transla-
. The bold lines indicate where the curvature κ has its first peak
The other point is the dependency of the estimate e
the trabecula was enclosed with a ty
between the FE analyses and the DVC are p
The more scattered dependency of the estimated e
combination ofvolume size of 40×40×40 voxels and a regularization factor of 100 was
sizes andhe mean and
m-a-
. The bold lines indicate where the curvature κ has its first peak
The other point is the dependency of the estimate
the trabecula was enclosed with a tythe trabecula. Therefore,
between the FE analyses and the DVC are p
The other point is the dependency of the estimated elastic modulus on thelength of the discretized trabeculae in the FE model. In the experiment, thetip of the trabecula sometimes deformed severely during compression. Addi-tionally, the other end of the trabecula was enclosed with a type of epoxyglue, which might deform differently from the trabecula. Therefore, bothends of the trabecula were gradually excluded from the FE model, with λdenoting the length ratio between the discretized trabeculae and the physicaltrabecula (Fig. 21). The relative difference in cross-sectional displacementbetween the FE analyses and the DVC are plotted for each cross sectionalong the longitudinal (trabecular length) direction, calculated as
= , (6)
where is the longitudinal displacement from FE analyses, is the longi-tudinal displacement from DVC, the superscript denotes all nodal values incross section in the two displacement fields.
Figure 22b shows that the relative difference δ decreased with decreasingλ, as the trabecular slenderness decreased, and more glue and damaged tipregions were excluded.
The more scattered dependency of the estimated elastic moduli on λ isshown in Fig. 22c. After excluding the results from FE models giving rela-
tive differenceavetrabecular samples. Thoughproposedof
Figurelength λL.
Figureeach cross section along the trabeculae for λ=0.2;0.4;0.6;0.8. c) estimated elasticmoduli at each λ. The black and white markers indicate which values were selectedfor continued anf
tive differenceaverage elastic modulus estimated watrabecular samples. Thoughproposedof single trabecula
Figurelength λL.
Figureeach cross section along the trabeculae for λ=0.2;0.4;0.6;0.8. c) estimated elasticmoduli at each λ. The black and white markers indicate which values were selectedfor continued anfrom Paper III.
tive differencerage elastic modulus estimated wa
trabecular samples. Thoughproposed
single trabecula
Figure 21length λL.
Figure 22each cross section along the trabeculae for λ=0.2;0.4;0.6;0.8. c) estimated elasticmoduli at each λ. The black and white markers indicate which values were selectedfor continued anrom Paper III.
tive differencerage elastic modulus estimated wa
trabecular samples. Thoughproposed method shows a
single trabecula
21. The physical trabecula of length L, and the discretized trabecula oflength λL. Reprinted
22. a) Trabecular shapes from initial CT scans; b) relative differences δ ofeach cross section along the trabeculae for λ=0.2;0.4;0.6;0.8. c) estimated elasticmoduli at each λ. The black and white markers indicate which values were selectedfor continued anrom Paper III.
tive difference δrage elastic modulus estimated wa
trabecular samples. Thoughmethod shows a
single trabecula
The physical trabecula of length L, and the discretized trabecula ofReprinted
a) Trabecular shapes from initial CT scans; b) relative differences δ ofeach cross section along the trabeculae for λ=0.2;0.4;0.6;0.8. c) estimated elasticmoduli at each λ. The black and white markers indicate which values were selectedfor continued analyses.rom Paper III.
δ largerrage elastic modulus estimated wa
trabecular samples. Thoughmethod shows a
single trabeculae
The physical trabecula of length L, and the discretized trabecula ofReprinted f
a) Trabecular shapes from initial CT scans; b) relative differences δ ofeach cross section along the trabeculae for λ=0.2;0.4;0.6;0.8. c) estimated elasticmoduli at each λ. The black and white markers indicate which values were selected
alyses.
argerrage elastic modulus estimated wa
trabecular samples. Thoughmethod shows a
e.
The physical trabecula of length L, and the discretized trabecula offrom Paper III.
a) Trabecular shapes from initial CT scans; b) relative differences δ ofeach cross section along the trabeculae for λ=0.2;0.4;0.6;0.8. c) estimated elasticmoduli at each λ. The black and white markers indicate which values were selected
alyses. X
arger than 10% for over 20% of the trabecular lengthrage elastic modulus estimated wa
trabecular samples. Thoughmethod shows a
The physical trabecula of length L, and the discretized trabecula ofrom Paper III.
a) Trabecular shapes from initial CT scans; b) relative differences δ ofeach cross section along the trabeculae for λ=0.2;0.4;0.6;0.8. c) estimated elasticmoduli at each λ. The black and white markers indicate which values were selected
X3 is the coordinate in the longitudinal direction.
than 10% for over 20% of the trabecular lengthrage elastic modulus estimated wa
trabecular samples. Though this study wasmethod shows a potential
The physical trabecula of length L, and the discretized trabecula ofrom Paper III.
a) Trabecular shapes from initial CT scans; b) relative differences δ ofeach cross section along the trabeculae for λ=0.2;0.4;0.6;0.8. c) estimated elasticmoduli at each λ. The black and white markers indicate which values were selected
is the coordinate in the longitudinal direction.
than 10% for over 20% of the trabecular lengthrage elastic modulus estimated wa
this study waspotential
The physical trabecula of length L, and the discretized trabecula ofrom Paper III.
a) Trabecular shapes from initial CT scans; b) relative differences δ ofeach cross section along the trabeculae for λ=0.2;0.4;0.6;0.8. c) estimated elasticmoduli at each λ. The black and white markers indicate which values were selected
is the coordinate in the longitudinal direction.
than 10% for over 20% of the trabecular lengthrage elastic modulus estimated wa
this study waspotential for
The physical trabecula of length L, and the discretized trabecula of
a) Trabecular shapes from initial CT scans; b) relative differences δ ofeach cross section along the trabeculae for λ=0.2;0.4;0.6;0.8. c) estimated elasticmoduli at each λ. The black and white markers indicate which values were selected
is the coordinate in the longitudinal direction.
than 10% for over 20% of the trabecular lengthrage elastic modulus estimated was 3.83 ± 0.54 GPa for three human
this study wasfor estimat
The physical trabecula of length L, and the discretized trabecula of
a) Trabecular shapes from initial CT scans; b) relative differences δ ofeach cross section along the trabeculae for λ=0.2;0.4;0.6;0.8. c) estimated elasticmoduli at each λ. The black and white markers indicate which values were selected
is the coordinate in the longitudinal direction.
than 10% for over 20% of the trabecular lengths 3.83 ± 0.54 GPa for three human
this study was limited by the sample size, theestimat
The physical trabecula of length L, and the discretized trabecula of
a) Trabecular shapes from initial CT scans; b) relative differences δ ofeach cross section along the trabeculae for λ=0.2;0.4;0.6;0.8. c) estimated elasticmoduli at each λ. The black and white markers indicate which values were selected
is the coordinate in the longitudinal direction.
than 10% for over 20% of the trabecular lengths 3.83 ± 0.54 GPa for three human
limited by the sample size, theestimating
The physical trabecula of length L, and the discretized trabecula of
a) Trabecular shapes from initial CT scans; b) relative differences δ ofeach cross section along the trabeculae for λ=0.2;0.4;0.6;0.8. c) estimated elasticmoduli at each λ. The black and white markers indicate which values were selected
is the coordinate in the longitudinal direction.
than 10% for over 20% of the trabecular lengths 3.83 ± 0.54 GPa for three human
limited by the sample size, theing the mechanical properties
The physical trabecula of length L, and the discretized trabecula of
a) Trabecular shapes from initial CT scans; b) relative differences δ ofeach cross section along the trabeculae for λ=0.2;0.4;0.6;0.8. c) estimated elasticmoduli at each λ. The black and white markers indicate which values were selected
is the coordinate in the longitudinal direction.
than 10% for over 20% of the trabecular lengths 3.83 ± 0.54 GPa for three human
limited by the sample size, thethe mechanical properties
The physical trabecula of length L, and the discretized trabecula of
a) Trabecular shapes from initial CT scans; b) relative differences δ ofeach cross section along the trabeculae for λ=0.2;0.4;0.6;0.8. c) estimated elasticmoduli at each λ. The black and white markers indicate which values were selected
is the coordinate in the longitudinal direction.
than 10% for over 20% of the trabecular lengths 3.83 ± 0.54 GPa for three human
limited by the sample size, thethe mechanical properties
The physical trabecula of length L, and the discretized trabecula of
a) Trabecular shapes from initial CT scans; b) relative differences δ ofeach cross section along the trabeculae for λ=0.2;0.4;0.6;0.8. c) estimated elasticmoduli at each λ. The black and white markers indicate which values were selected
is the coordinate in the longitudinal direction.
than 10% for over 20% of the trabecular lengths 3.83 ± 0.54 GPa for three human
limited by the sample size, thethe mechanical properties
The physical trabecula of length L, and the discretized trabecula of
a) Trabecular shapes from initial CT scans; b) relative differences δ ofeach cross section along the trabeculae for λ=0.2;0.4;0.6;0.8. c) estimated elasticmoduli at each λ. The black and white markers indicate which values were selected
is the coordinate in the longitudinal direction.
than 10% for over 20% of the trabecular lengths 3.83 ± 0.54 GPa for three human
limited by the sample size, thethe mechanical properties
The physical trabecula of length L, and the discretized trabecula of
a) Trabecular shapes from initial CT scans; b) relative differences δ ofeach cross section along the trabeculae for λ=0.2;0.4;0.6;0.8. c) estimated elasticmoduli at each λ. The black and white markers indicate which values were selected
is the coordinate in the longitudinal direction.
than 10% for over 20% of the trabecular lengths 3.83 ± 0.54 GPa for three human
limited by the sample size, thethe mechanical properties
The physical trabecula of length L, and the discretized trabecula of
a) Trabecular shapes from initial CT scans; b) relative differences δ ofeach cross section along the trabeculae for λ=0.2;0.4;0.6;0.8. c) estimated elasticmoduli at each λ. The black and white markers indicate which values were selected
Reprinted
41
than 10% for over 20% of the trabecular length, thes 3.83 ± 0.54 GPa for three human
limited by the sample size, thethe mechanical properties
The physical trabecula of length L, and the discretized trabecula of
a) Trabecular shapes from initial CT scans; b) relative differences δ ofeach cross section along the trabeculae for λ=0.2;0.4;0.6;0.8. c) estimated elasticmoduli at each λ. The black and white markers indicate which values were selected
Reprinted
41
, thes 3.83 ± 0.54 GPa for three human
limited by the sample size, thethe mechanical properties
The physical trabecula of length L, and the discretized trabecula of
a) Trabecular shapes from initial CT scans; b) relative differences δ ofeach cross section along the trabeculae for λ=0.2;0.4;0.6;0.8. c) estimated elasticmoduli at each λ. The black and white markers indicate which values were selected
Reprinted
42
Paper IV - Quantification of strains and cracksin trabecular bone by digital volumecorrelation: A case study
The combination of step-wise loading, micro-CT and DVC has enabled thevisualization and quantification of internal deformations of trabecular struc-tures. The local strain fields can provide a ‘cross-validation’ for finite ele-ment models of these structures, which have been widely applied to simulatebone mechanical behavior [93-95]. Whilst deformation, cracks and damageaccumulation have been observed locally on individual trabeculae in the teststructures, strains have seldom been estimated at this level due to relativelylow image resolutions (20-40 µm) and therefore too low DVC spatial resolu-tions (0.8-2 mm) to investigate single trabeculae. Recently, local strains wereestimated by DVC on high-resolution images from SRµCT of trabecularbone loaded in the apparent elastic range [75]. The aim of this case studywas to quantify trabecular strains and cracks during compression until failureusing the proposed DVC algorithm (Paper II) on SRµCT images.
An in-situ compression test was performed on a human trabecular cylin-der, which was scanned at a voxel size of 3.25 µm. The sub-volume size andregularization factor of the DVC method, 64 and 60 respectively, were se-lected based on a test-volume of 200×200×200 voxels using the same strate-gy as discussed in Paper III, i.e. determined through a balance of low virtualstrain energy and high accuracy.
Figure 23 shows a general increase in the first principle strains, and high-er strains near the convex surface where displacement was gradually applied(Fig. 23). Strain concentrations were observed in trabeculae at a distancefrom the top, indicating an effect of the discrete nature of the trabecular ar-chitecture to the strain distribution. When zoomed in to the highly deformedtrabeculae, we noticed that individual trabeculae were subjected to high ten-sile and shear strains though the trabecular bone was under compression atthe apparent level. Predominant open mode fractures and some few shearmode fractures were found by contrasting the cracks and tensile/shearstrains.
4343
Figure 23. The first principle strain (EI) of the trabecular specimen at loading steps2-5. Displacement was applied in the negative Z direction. Strain was plotted in theinitial configurations. Reproduced from Paper IV.
Cracks in the trabecular bone were segmented with DVC-based image processing. Based on step 6 where an obvious increase in crack volume was noticed (Fig. 24), the trabecular structure was divided into cracked and ‘uncracked’ regions. Mean and standard deviations of the first principal strains in the two groups were contrasted at steps 2-6. Regions in the trabecu-lae that had visible cracks in step 6 showed higher strains than the uncracked regions. The difference between the two groups increased slightly from step 2 to step 5, and was a factor of 7.6 at step 6. When the cracks in step 6 were grouped by the crack volume, larger opening cracks showed higher tensile strains than smaller cracks. Note that not all regions with high strains resulted in cracks and some trabeculae with low strain in step 5 were observed with cracks in step 6.
Figure 24. Load and displacement at the load steps in the uniaxial compressiontesting, as well as crack volumes estimated by a DVC-based image processing andnormalized by the total bone volume (left). Mean and standard deviations of the firstprincipal strain at steps 2-6 in the cracked and ‘uncracked’ regions. Reproducedfrom Paper IV.
44
Paper V - The effect of two augmentationmaterials on screw pullout resistance fromhuman trabecular bone
Augmentation materials, such as CPCs and PMMA, have been used to im-prove the physical engagement of screws inserted into bone. While ceramic,degradable cements may ultimately improve the fixation [49, 51, 96], incon-sistent effects on early stability have been reported [57].
A recently developed degradable ceramic adhesive has been reported tobe able to bond with tissues with relatively high strength [60]. Consideringthat the bonding of screws with surrounding bone may contribute to moreevenly distributed loads in the vicinity of the screw and along the threads,augmentation with the adhesive may improve the pullout performance. Thepullout resistance may be further enhanced by transiting to adhesive bondingfrom physical contact at the interfaces, with improved shear and tensile frac-ture toughness. The aim of this study was to investigate the screw pulloutresistance and failure mechanisms from trabecular bone, and the effect oftwo augmentation materials, i.e. a ceramic cement or an adhesive material,with no augmentation as a control.
Two pullout tests were performed, one of which was an ex-situ continu-ous pullout of screws from 21 trabecular specimens which were divided intothree groups by the augmentation materials, i.e. cement, adhesive and noaugmentation as a control (Fig. 25). The other one was an in-situ step-wisepullout inside a SRμCT. The five specimens for the in-situ testing werenamed following the form: augmentation material-in-specimen number.
45
Figure 25. Section views from micro-CT scans showing the inserted screws and thedistribution of the augmentation materials in ex-situ specimens in three groups: (a)control, (b) cement, (c) adhesive. Reprinted from Paper V.
Two parameters were characterized from the micro-CT images: BV/TV and injection volume of the augmentation materials. BV/TV was calculated from the images acquired before the screw insertion. The injection volume was estimated based on the comparison of micro-CT images before and after the screw insertion.
Linear regression showed the highest correlation of the pullout load withBV/TV in the region just around the screw. The high correlation was in ac-cordance with the earlier reported main effect of BV/TV [18, 97] or bonemineral density [98, 99] on the pullout strength, confirming the contributionof trabecular bone in the direct vicinity of the screw to the pullout resistance.
The pullout load, stiffness and fracture energy (calculated as the area un-der the load-displacement curve until maximum load) of the ex-situ testing normalized with BV/TV were compared among the three groups by one-way ANOVA (Fig. 26). Augmentation with the cement did not affect the pullout resistance, compared to the control group. In contrast, the pullout resistance after augmentation with the adhesive was significantly higher than the cement and the control group. This enhancement is comparable to the literature-reported enhancement after augmentation with ceramic cement, but slightly worse than with PMMA in vertebrae (1.8-2.8 fold [49, 56, 100]).
Note that the injection volume and the distribution of the two augmenta-tion material were different (Fig. 25), which might affect the pullout re-sistance. When the pullout properties were normalized with a factor thatincluded both BV/TV and injection volume, there were no significant differ-ences in most cases. However, the linear relation between pullout propertieswith this factor needs to be verified. On the other hand, the adhesive showedan advantage in injectability and material distribution, compared to thebrushite cement used.
46
Figurefracture energy of the exing for BV/TV. The asterisk mark (*) denotes significant difference between twocompared groups.
ing steps are exemplified for one specimen with/without augmentation (27propagate[101]augmented with the cementadhesivetrabeculaand theperipheries of thethe adhesive and trabeculae near the screw.
FigureControlplacement in mm
46
Figurefracture energy of the exing for BV/TV. The asterisk mark (*) denotes significant difference between twocompared groups.
Slices of the periing steps are exemplified for one specimen with/without augmentation (27-30propagate[101]augmented with the cementadhesivetrabeculaand theperipheries of thethe adhesive and trabeculae near the screw.
FigureControlplacement in mm
Figure 26fracture energy of the exing for BV/TV. The asterisk mark (*) denotes significant difference between twocompared groups.
Slices of the periing steps are exemplified for one specimen with/without augmentation (
30). In a specimen without augmentationpropagate[101]. Laugmented with the cementadhesivetrabeculaand theperipheries of thethe adhesive and trabeculae near the screw.
Figure 27Control-inplacement in mm
26. Estimated marginfracture energy of the exing for BV/TV. The asterisk mark (*) denotes significant difference between twocompared groups.
Slices of the periing steps are exemplified for one specimen with/without augmentation (
). In a specimen without augmentationpropagated
Long cracks across several trabeculae were not observedaugmented with the cementadhesive (Adhesietrabeculae, and cracks were observed to propagate through the trabeculaeand the adhesperipheries of thethe adhesive and trabeculae near the screw.
27. Crack propagation across several trabeculae during screw pullout fromin-1. ‘B’ denotes bone, ‘S’ denotes the screw
placement in mm
Estimated marginfracture energy of the exing for BV/TV. The asterisk mark (*) denotes significant difference between twocompared groups.
Slices of the periing steps are exemplified for one specimen with/without augmentation (
). In a specimen without augmentationacross the trabeculae, similar to the crack paths in brittle foams
ong cracks across several trabeculae were not observedaugmented with the cement
(Adhesie, and cracks were observed to propagate through the trabeculae
adhesive. These cracks initiated at both the thread crests and theperipheries of thethe adhesive and trabeculae near the screw.
Crack propagation across several trabeculae during screw pullout from1. ‘B’ denotes bone, ‘S’ denotes the screw
placement in mm
Estimated marginfracture energy of the exing for BV/TV. The asterisk mark (*) denotes significant difference between twocompared groups. Reprinted
Slices of the periing steps are exemplified for one specimen with/without augmentation (
). In a specimen without augmentationacross the trabeculae, similar to the crack paths in brittle foams
ong cracks across several trabeculae were not observedaugmented with the cement
(Adhesie-, and cracks were observed to propagate through the trabeculae
ive. These cracks initiated at both the thread crests and theperipheries of the adhesive, indicating an improved load distribution thrthe adhesive and trabeculae near the screw.
Crack propagation across several trabeculae during screw pullout from1. ‘B’ denotes bone, ‘S’ denotes the screw
placement in mm. Reprinted
Estimated marginfracture energy of the ex-ing for BV/TV. The asterisk mark (*) denotes significant difference between two
Reprinted
Slices of the peri-implant region from the ining steps are exemplified for one specimen with/without augmentation (
). In a specimen without augmentationacross the trabeculae, similar to the crack paths in brittle foams
ong cracks across several trabeculae were not observedaugmented with the cement
-in-1,, and cracks were observed to propagate through the trabeculae
ive. These cracks initiated at both the thread crests and theadhesive, indicating an improved load distribution thr
the adhesive and trabeculae near the screw.
Crack propagation across several trabeculae during screw pullout from1. ‘B’ denotes bone, ‘S’ denotes the screw
Reprinted
Estimated margin-situ specimens, given by the general linear model accoun
ing for BV/TV. The asterisk mark (*) denotes significant difference between twoReprinted from Paper
implant region from the ining steps are exemplified for one specimen with/without augmentation (
). In a specimen without augmentationacross the trabeculae, similar to the crack paths in brittle foams
ong cracks across several trabeculae were not observedaugmented with the cement
1, Fig, and cracks were observed to propagate through the trabeculae
ive. These cracks initiated at both the thread crests and theadhesive, indicating an improved load distribution thr
the adhesive and trabeculae near the screw.
Crack propagation across several trabeculae during screw pullout from1. ‘B’ denotes bone, ‘S’ denotes the screw
Reprinted from Paper V.
Estimated marginal means of the (a) pullout load, (b) stiffness and (c)situ specimens, given by the general linear model accoun
ing for BV/TV. The asterisk mark (*) denotes significant difference between twofrom Paper
implant region from the ining steps are exemplified for one specimen with/without augmentation (
). In a specimen without augmentationacross the trabeculae, similar to the crack paths in brittle foams
ong cracks across several trabeculae were not observedaugmented with the cement. However, in the specimen augmented with the
Figs., and cracks were observed to propagate through the trabeculae
ive. These cracks initiated at both the thread crests and theadhesive, indicating an improved load distribution thr
the adhesive and trabeculae near the screw.
Crack propagation across several trabeculae during screw pullout from1. ‘B’ denotes bone, ‘S’ denotes the screw
from Paper V.
al means of the (a) pullout load, (b) stiffness and (c)situ specimens, given by the general linear model accoun
ing for BV/TV. The asterisk mark (*) denotes significant difference between twofrom Paper
implant region from the ining steps are exemplified for one specimen with/without augmentation (
). In a specimen without augmentationacross the trabeculae, similar to the crack paths in brittle foams
ong cracks across several trabeculae were not observed. However, in the specimen augmented with the
27-28, and cracks were observed to propagate through the trabeculae
ive. These cracks initiated at both the thread crests and theadhesive, indicating an improved load distribution thr
the adhesive and trabeculae near the screw.
Crack propagation across several trabeculae during screw pullout from1. ‘B’ denotes bone, ‘S’ denotes the screw
from Paper V.
al means of the (a) pullout load, (b) stiffness and (c)situ specimens, given by the general linear model accoun
ing for BV/TV. The asterisk mark (*) denotes significant difference between twofrom Paper V.
implant region from the ining steps are exemplified for one specimen with/without augmentation (
). In a specimen without augmentationacross the trabeculae, similar to the crack paths in brittle foams
ong cracks across several trabeculae were not observed. However, in the specimen augmented with the
28), the, and cracks were observed to propagate through the trabeculae
ive. These cracks initiated at both the thread crests and theadhesive, indicating an improved load distribution thr
the adhesive and trabeculae near the screw.
Crack propagation across several trabeculae during screw pullout from1. ‘B’ denotes bone, ‘S’ denotes the screw
from Paper V.
al means of the (a) pullout load, (b) stiffness and (c)situ specimens, given by the general linear model accoun
ing for BV/TV. The asterisk mark (*) denotes significant difference between twoV.
implant region from the ining steps are exemplified for one specimen with/without augmentation (
). In a specimen without augmentationacross the trabeculae, similar to the crack paths in brittle foams
ong cracks across several trabeculae were not observed. However, in the specimen augmented with the
, the, and cracks were observed to propagate through the trabeculae
ive. These cracks initiated at both the thread crests and theadhesive, indicating an improved load distribution thr
the adhesive and trabeculae near the screw.
Crack propagation across several trabeculae during screw pullout from1. ‘B’ denotes bone, ‘S’ denotes the screw
from Paper V.
al means of the (a) pullout load, (b) stiffness and (c)situ specimens, given by the general linear model accoun
ing for BV/TV. The asterisk mark (*) denotes significant difference between two
implant region from the ining steps are exemplified for one specimen with/without augmentation (
). In a specimen without augmentationacross the trabeculae, similar to the crack paths in brittle foams
ong cracks across several trabeculae were not observed. However, in the specimen augmented with the
, the adhesive filled the space between, and cracks were observed to propagate through the trabeculae
ive. These cracks initiated at both the thread crests and theadhesive, indicating an improved load distribution thr
the adhesive and trabeculae near the screw.
Crack propagation across several trabeculae during screw pullout from1. ‘B’ denotes bone, ‘S’ denotes the screw
al means of the (a) pullout load, (b) stiffness and (c)situ specimens, given by the general linear model accoun
ing for BV/TV. The asterisk mark (*) denotes significant difference between two
implant region from the in-situ testing at different loaing steps are exemplified for one specimen with/without augmentation (
). In a specimen without augmentation (Controlacross the trabeculae, similar to the crack paths in brittle foams
ong cracks across several trabeculae were not observed. However, in the specimen augmented with the
adhesive filled the space between, and cracks were observed to propagate through the trabeculae
ive. These cracks initiated at both the thread crests and theadhesive, indicating an improved load distribution thr
Crack propagation across several trabeculae during screw pullout from1. ‘B’ denotes bone, ‘S’ denotes the screw, and ‘d’ denotes applied di
al means of the (a) pullout load, (b) stiffness and (c)situ specimens, given by the general linear model accoun
ing for BV/TV. The asterisk mark (*) denotes significant difference between two
situ testing at different loaing steps are exemplified for one specimen with/without augmentation (
(Controlacross the trabeculae, similar to the crack paths in brittle foams
ong cracks across several trabeculae were not observed. However, in the specimen augmented with the
adhesive filled the space between, and cracks were observed to propagate through the trabeculae
ive. These cracks initiated at both the thread crests and theadhesive, indicating an improved load distribution thr
Crack propagation across several trabeculae during screw pullout from, and ‘d’ denotes applied di
al means of the (a) pullout load, (b) stiffness and (c)situ specimens, given by the general linear model accoun
ing for BV/TV. The asterisk mark (*) denotes significant difference between two
situ testing at different loaing steps are exemplified for one specimen with/without augmentation (
(Control-inacross the trabeculae, similar to the crack paths in brittle foams
ong cracks across several trabeculae were not observed. However, in the specimen augmented with the
adhesive filled the space between, and cracks were observed to propagate through the trabeculae
ive. These cracks initiated at both the thread crests and theadhesive, indicating an improved load distribution thr
Crack propagation across several trabeculae during screw pullout from, and ‘d’ denotes applied di
al means of the (a) pullout load, (b) stiffness and (c)situ specimens, given by the general linear model accoun
ing for BV/TV. The asterisk mark (*) denotes significant difference between two
situ testing at different loaing steps are exemplified for one specimen with/without augmentation (
in-1,across the trabeculae, similar to the crack paths in brittle foams
ong cracks across several trabeculae were not observed. However, in the specimen augmented with the
adhesive filled the space between, and cracks were observed to propagate through the trabeculae
ive. These cracks initiated at both the thread crests and theadhesive, indicating an improved load distribution thr
Crack propagation across several trabeculae during screw pullout from, and ‘d’ denotes applied di
al means of the (a) pullout load, (b) stiffness and (c)situ specimens, given by the general linear model accoun
ing for BV/TV. The asterisk mark (*) denotes significant difference between two
situ testing at different loaing steps are exemplified for one specimen with/without augmentation (
Fig.across the trabeculae, similar to the crack paths in brittle foams
ong cracks across several trabeculae were not observed. However, in the specimen augmented with the
adhesive filled the space between, and cracks were observed to propagate through the trabeculae
ive. These cracks initiated at both the thread crests and theadhesive, indicating an improved load distribution thr
Crack propagation across several trabeculae during screw pullout from, and ‘d’ denotes applied di
al means of the (a) pullout load, (b) stiffness and (c)situ specimens, given by the general linear model accoun
ing for BV/TV. The asterisk mark (*) denotes significant difference between two
situ testing at different loaing steps are exemplified for one specimen with/without augmentation (
Fig. 27across the trabeculae, similar to the crack paths in brittle foams
in specimens. However, in the specimen augmented with the
adhesive filled the space between, and cracks were observed to propagate through the trabeculae
ive. These cracks initiated at both the thread crests and theadhesive, indicating an improved load distribution thr
Crack propagation across several trabeculae during screw pullout from, and ‘d’ denotes applied di
al means of the (a) pullout load, (b) stiffness and (c)situ specimens, given by the general linear model accoun
ing for BV/TV. The asterisk mark (*) denotes significant difference between two
situ testing at different loaing steps are exemplified for one specimen with/without augmentation (
27), cracksacross the trabeculae, similar to the crack paths in brittle foams
in specimens. However, in the specimen augmented with the
adhesive filled the space between, and cracks were observed to propagate through the trabeculae
ive. These cracks initiated at both the thread crests and theadhesive, indicating an improved load distribution thr
Crack propagation across several trabeculae during screw pullout from, and ‘d’ denotes applied di
al means of the (a) pullout load, (b) stiffness and (c)situ specimens, given by the general linear model accoun
ing for BV/TV. The asterisk mark (*) denotes significant difference between two
situ testing at different load-ing steps are exemplified for one specimen with/without augmentation (Figs.
, cracksacross the trabeculae, similar to the crack paths in brittle foams
in specimens. However, in the specimen augmented with the
adhesive filled the space between, and cracks were observed to propagate through the trabeculae
ive. These cracks initiated at both the thread crests and theadhesive, indicating an improved load distribution through
Crack propagation across several trabeculae during screw pullout from, and ‘d’ denotes applied dis-
al means of the (a) pullout load, (b) stiffness and (c)situ specimens, given by the general linear model account-
ing for BV/TV. The asterisk mark (*) denotes significant difference between two
d-Figs.
, cracksacross the trabeculae, similar to the crack paths in brittle foams
in specimens. However, in the specimen augmented with the
adhesive filled the space between, and cracks were observed to propagate through the trabeculae
ive. These cracks initiated at both the thread crests and theough
Crack propagation across several trabeculae during screw pullout from
Figurefrom Adhesive‘d’ denotes applied displacement in mm
Figuresiveand propagated towards the screw. ‘A’ denotesdenotes screw V.
Figurefrom Adhesive‘d’ denotes applied displacement in mm
Figuresive-and propagated towards the screw. ‘A’ denotesdenotes screw V.
Figure 29from Adhesive‘d’ denotes applied displacement in mm
Figure 30-in-1, showing crack initiated at the periphery of the bone
and propagated towards the screw. ‘A’ denotesdenotes screw
29. Crack propagation across trabeculae and adhesive during screw pulloutfrom Adhesive‘d’ denotes applied displacement in mm
30. Induced cracks in bone and adhesive during screw pullout from Adh1, showing crack initiated at the periphery of the bone
and propagated towards the screw. ‘A’ denotesdenotes screw
Crack propagation across trabeculae and adhesive during screw pulloutfrom Adhesive-in‘d’ denotes applied displacement in mm
Induced cracks in bone and adhesive during screw pullout from Adh1, showing crack initiated at the periphery of the bone
and propagated towards the screw. ‘A’ denotesdenotes screw, and ‘d’ denot
Crack propagation across trabeculae and adhesive during screw pulloutin-1. ‘A’ deno
‘d’ denotes applied displacement in mm
Induced cracks in bone and adhesive during screw pullout from Adh1, showing crack initiated at the periphery of the bone
and propagated towards the screw. ‘A’ denotes, and ‘d’ denot
Crack propagation across trabeculae and adhesive during screw pullout1. ‘A’ deno
‘d’ denotes applied displacement in mm
Induced cracks in bone and adhesive during screw pullout from Adh1, showing crack initiated at the periphery of the bone
and propagated towards the screw. ‘A’ denotes, and ‘d’ denot
Crack propagation across trabeculae and adhesive during screw pullout1. ‘A’ deno
‘d’ denotes applied displacement in mm
Induced cracks in bone and adhesive during screw pullout from Adh1, showing crack initiated at the periphery of the bone
and propagated towards the screw. ‘A’ denotes, and ‘d’ denot
Crack propagation across trabeculae and adhesive during screw pullout1. ‘A’ denotes adhesive, ‘B’ denotes bone,
‘d’ denotes applied displacement in mm
Induced cracks in bone and adhesive during screw pullout from Adh1, showing crack initiated at the periphery of the bone
and propagated towards the screw. ‘A’ denotes, and ‘d’ denotes applied displacement in mm
Crack propagation across trabeculae and adhesive during screw pullouttes adhesive, ‘B’ denotes bone,
‘d’ denotes applied displacement in mm
Induced cracks in bone and adhesive during screw pullout from Adh1, showing crack initiated at the periphery of the bone
and propagated towards the screw. ‘A’ denoteses applied displacement in mm
Crack propagation across trabeculae and adhesive during screw pullouttes adhesive, ‘B’ denotes bone,
‘d’ denotes applied displacement in mm. Reprinted
Induced cracks in bone and adhesive during screw pullout from Adh1, showing crack initiated at the periphery of the bone
and propagated towards the screw. ‘A’ denoteses applied displacement in mm
Crack propagation across trabeculae and adhesive during screw pullouttes adhesive, ‘B’ denotes bone,
Reprinted
Induced cracks in bone and adhesive during screw pullout from Adh1, showing crack initiated at the periphery of the bone
and propagated towards the screw. ‘A’ denoteses applied displacement in mm
Crack propagation across trabeculae and adhesive during screw pullouttes adhesive, ‘B’ denotes bone,
Reprinted
Induced cracks in bone and adhesive during screw pullout from Adh1, showing crack initiated at the periphery of the bone
and propagated towards the screw. ‘A’ denoteses applied displacement in mm
Crack propagation across trabeculae and adhesive during screw pullouttes adhesive, ‘B’ denotes bone,
Reprinted from Paper V.
Induced cracks in bone and adhesive during screw pullout from Adh1, showing crack initiated at the periphery of the bone
and propagated towards the screw. ‘A’ denotes adhesive, ‘B’ denotes bone,es applied displacement in mm
Crack propagation across trabeculae and adhesive during screw pullouttes adhesive, ‘B’ denotes bone,
from Paper V.
Induced cracks in bone and adhesive during screw pullout from Adh1, showing crack initiated at the periphery of the bone
adhesive, ‘B’ denotes bone,es applied displacement in mm
Crack propagation across trabeculae and adhesive during screw pullouttes adhesive, ‘B’ denotes bone, ‘S’ denotes screw
from Paper V.
Induced cracks in bone and adhesive during screw pullout from Adh1, showing crack initiated at the periphery of the bone
adhesive, ‘B’ denotes bone,es applied displacement in mm.
Crack propagation across trabeculae and adhesive during screw pullout‘S’ denotes screw
from Paper V.
Induced cracks in bone and adhesive during screw pullout from Adh1, showing crack initiated at the periphery of the bone-adhesive composite
adhesive, ‘B’ denotes bone,Reprinted
Crack propagation across trabeculae and adhesive during screw pullout‘S’ denotes screw
Induced cracks in bone and adhesive during screw pullout from Adhadhesive composite
adhesive, ‘B’ denotes bone,Reprinted
Crack propagation across trabeculae and adhesive during screw pullout‘S’ denotes screw
Induced cracks in bone and adhesive during screw pullout from Adhadhesive composite
adhesive, ‘B’ denotes bone,Reprinted from Paper
Crack propagation across trabeculae and adhesive during screw pullout‘S’ denotes screw
Induced cracks in bone and adhesive during screw pullout from Adhadhesive composite
adhesive, ‘B’ denotes bone,from Paper
47
Crack propagation across trabeculae and adhesive during screw pullout‘S’ denotes screw, and
Induced cracks in bone and adhesive during screw pullout from Adhadhesive composite
adhesive, ‘B’ denotes bone, ‘S’from Paper
47
Crack propagation across trabeculae and adhesive during screw pullout, and
Induced cracks in bone and adhesive during screw pullout from Adhe-adhesive composite
‘S’from Paper
48
Paper VI - 3D-printed PLA/HA compositestructures as synthetic trabecular bone: afeasibility study using Fused DepositionModelling
Synthetic models mimicking trabecular architectures may be beneficial forunderstanding the mechanical effects of the trabecular structure in the evalu-ation of implant design for example. The flexibility of 3D-printing tech-niques, together with micro-CT, provides a possibility to reproduce a specif-ic trabecular bone. The aim of this study was to evaluate the feasibility ofprinting model trabecular bone using FDM and PLA/HA composites.
Micro-CT images of trabecular bone taken from a human femoral head were used to generate models for printing, and scaled synthetic bone was printed with PLA/HA composite filaments of three different HA concentra-tions, 5-10-15 wt%. Material characterization and compressive testing of the composite bulk material were performed. Morphology and mechanical prop-erties of the printed models were evaluated with micro-CT, compression and screw pullout tests.
Morphometric analyses showed that the printed PLA model could reproduce profiles of the input trabecular model with accuracy, albeit with significantly higher surface-to-volume ratio and lower trabecular thickness than the input model due to under-filling. In addition, incorpora-tion of HA particles reduced the printing accuracy.
Compression tests showed a significantly increased elastic modulus of theprinted bulk material with the incorporation of HA particles compared topure PLA. No significant difference in compressive yield strength wasshown between PLA with and without HA particles.
The printed trabecular bone showed higher variations in mechanical prop-erties compared to the bulk materials, but higher pullout load than a com-monly used synthetic model when scaled 4.3-fold in printing.
Though further work is needed to increase resolution, especially whenprinting with composite filaments, synthetic models mimicking trabecularstructure with PLA/HA composites -both components are biocompatible andresorbable- using FDM is a promising strategy to develop trabecular modelsfor mechanical evaluation, or potentially clinical applications.
49
6. Summary and conclusions
The high prevalence of bone fractures affects the quality of life of manypatients. Complications appear in the surgical treatment of bone fractureswith or without the use of bone substitutes or implants. Due to the fact thatmechanics is highly involved in the locomotion and that bone remodelingresponds to mechanical loading, the mechanical analyses of bone and itsinteraction with implants have to be better understood. The trabecular bonewhich is characterized by its spongy structure is a focus here. As a powerfultool to visualize and quantify the micro-level structural heterogeneity of thetrabecular bone, micro-CT has provided a base for mechanical related anal-yses.
This thesis has discussed several aspects on the topic of mechanical anal-yses of trabecular bone and its interaction with implants based on the en-closed papers. Brief summaries and main conclusions of these papers are listed below:
(1) The currently applied approaches to estimate the elastic modulus oftrabecular bone were reviewed, with highlights on important considerationsof these approaches, and values reported in the literature were summarized.
(2) A global DVC technique was proposed to estimate displacement andstrain fields based on 3D images. The displacement field was discretizedusing higher-order finite elements and solved by a global optimization pro-cedure. Compared to a commercial code, the proposed technique performedsomewhat better in most cases, showing a potential to estimate the displace-ment field with accuracy. This study also highlighted the need for codes tobe applied to double scans, virtual deformation and real deformation testcases when evaluating their performance.
(3) A method was developed to estimate the elastic modulus of single tra-beculae from compression testing. A global DVC algorithm was applied to estimate full-field displacements of the single trabeculae, based on high-resolution 3D images obtained by SRμCT. The displacement provided boundary conditions for high-resolution finite element models considering the irregular geometry of the trabeculae. The proposed method shows a po-tential to estimate single-trabeculae-level mechanical properties, while some possible method improvements were also identified.
(4) Strain fields and cracks of trabecular bone loaded to failure by step-wise compression were estimated by applying DVC on SRµCT images ac-quired during loading. Accumulation of strains and cracks at the individual
50
trabecular level were characterized. Trabeculae that were subsequently cracked showed higher strains than regions without segmented cracks during the loading by superimposing the cracks with strain maps. A potential for quantifying individual trabecular deformation by DVC on high-resolution im-ages was shown with this case study.
(5) Screw pullout resistance from trabecular bone with or without aug-mentation with a brushite cement, or a ceramic adhesive was compared. Theenhanced pullout resistance after augmentation with the adhesive, specifical-ly, higher pullout loads, stiffness, fracture energy and more distributedcracks, compared to the cement, indicated a promising augmentation materi-al for the primary implant stability. Cracks at the peripheries of the bone-adhesive composite indicate an observable connection between the threecomponents.
(6) Synthetic trabecular structures were printed using FDM and PLA/HA composite materials. The reproducibility in terms of morphology and increased pullout strength of the printed model indicate a potential for achieving better synthetic models compared to commonly used such, although resolu-tion needs to be improved.
The findings in this thesis improve and may contribute to further im-provements in the mechanical aspects of trabecular bone and implants, in-cluding but not limited to, mechanical properties of bone, augmentationstrategies and synthetic trabecular models for mechanical evaluations.
51
7. Future perspectives
Quantitative tomography, imaging analyses and numerical analyses are toolswidely used in tissue related studies; some have even been applied in clinicsfor decades. A combination of these tools with mechanical tests shows agreat potential to analyze the mechanical behavior of materials, especiallyheterogeneous such. Analysis of tomographic images can allow for visuali-zation and quantification of internal deformations. Results from numericalmodels can be compared with image analyses to improve the models or es-timate material mechanical properties. It is expected that these tools will beapplied for the analyses of a wider range of materials, and continuous im-provements in the accuracy or efficiency of these techniques are of greatbenefit. Regarding the topics of this thesis, further investigations or im-provements can be performed, for instance:
· improve the efficiency of the code by programming with a lower-level language;
· a reliable finite element model of screw pullout from trabecular bonemay be proposed by numerical modelling with boundary conditionsestimated by DVC and evaluation of the results based on DVCstrains and the experimentally observed cracks;
· the effect of bonding strength at the interfaces between screwthreads and the augmentation materials on the pullout resistance maybe further investigated by finite element analyses.
· evaluation of strain concentrations further away from the appliedload in the trabecular bone, and quantification of the distribution ofstrain concentrations and crack volume in the bone.
52
Svensk sammanfattning
Benfrakturer påverkar många patienters livskvalitet. Komplikationer upp-kommer ofta i samband med frakturrelaterad kirurgisk behandling, både medoch utan benersättningsmaterial eller implantat. Genom att vi rör oss påver-kas ben mekaniskt, och mekanisk last spelar en stor roll för benets remodel-lering. Det är därför viktigt att öka kunskapen om mekaniska aspekter avben, och benets interaktion med implantat, speciellt relaterat till porös ben-struktur (trabekulärt ben). Den här avhandlingen syftar till att undersökatrabekulärt bens mekaniska egenskaper genom analys av isolerade trabekler,hur benersättningsmaterial påverkar implantats primära stabilitet, och till attframställa syntetiska trabekulära modeller för experimentell mekanisk utvär-dering.
En litteraturstudie genomfördes för att ta reda på hur elasticitetsmodulen itrabekulärt ben på vävnadsnivå har undersökts tidigare. I studien identifiera-des fördelar och begränsningar med existerande metoder (Artikel I).
Artikel II utvecklade en global DVC-metod, baserad på diskretiseringmed högre ordningens finita element, för att uppskatta deformation i 3D-bilder. Den utvecklade metoden jämfördes med en annan global DVC-metodgenom både fysisk och virtuell deformation av bilder på trabekulärt ben medolika upplösning. I de flesta fallen presterade den nya tekniken bättre, och atttesta DVC-metoder med flertalet deformationsfält visades vara betydelse-fullt.
I Artikel III utfördes mekaniska tester in-situ på isolerade trabekler frånmänskligt ben under kompression i en SRµCT. Vidare utvecklades en metoddär DVC och FE analys användes för att beräkna elasticitetsmodulen. DVCutfördes på de högupplösta SRµCT-bilderna, och det resulterande deformat-ionsfältet kunde användas som randvillkor till FE modellen av isoleradetrabekler. Metoden visade sig ha potential för utvärdering av mekaniskaegenskaper av isolerade trabekler trots att några potentiella förbättringsmöj-ligheter identifierades.
I Artikel IV belastades trabekulärt ben till brott genom stegvis belastning ikompression. Deformationsfält och sprickor studerades på trabekulär nivågenom att applicera DVC på SRµCT-bilder tagna under belastning. En jäm-förelse mellan deformationsfälten och sprickorna från bilderna visade atttrabekler som gick sönder hade högre deformation i tidigare belastningsstegän regioner utan sprickor. Tekniken som presenteras i denna studie visar
53
potential för kvantifiering av deformation i individuella trabekler genomDVC på högupplösta bilder.
Artikel V undersökte hur förstärkning, eller augmentering, av ben medolika material påverkar den primära stabiliteten för ortopediska skruvar itrabekulärt ben. I de mekaniska testerna drogs skruvar ut från mänskligt tra-bekulärt ben, augmenterat med kalciumfosfatbaserat benlim eller kalcium-fosfat-cement, och icke-augmenterat ben användes som kontroll. En del avtesterna utfördes in situ i SRµCT. Ex situ-testerna visade att augmenteringmed benlim gav ett ökat motstånd, och från in situ-bilderna observerades detatt sprickorna var mer jämt distribuerade jämfört med i de andra grupperna.Resultaten visar benlimmets potential för augmentering av skruvar i trabeku-lärt ben.
I Artikel VI utvärderades möjligheten att 3D-printa en trabekulär strukturav PLA/HA-komposit genom FDM-teknik. I studien togs ett första steg föratt framställa syntetiska modeller som efterliknar trabekulär benstruktur,men upplösningen som uppnåddes i studien behöver förbättras.
Med den här avhandlingen förbättras kunskapen om mekaniska aspekter av trabekulärt ben och benets interaktion med implantat. Resultaten bidrar förbättringar i områden relaterade till mekaniska egenskaper av ben, strategier vid augmentering av ortopediska skruvar och framställning av syntetiska benmodeller för experimentell mekanisk utvärdering.
54
Acknowledgements
First and foremost, I would like to express my deep and sincere gratitude tomy supervisors, Prof. Cecilia Persson, and Prof. Per Isaksson for giving methe opportunity to carry out studies on this interesting topic. Cecilia, thankyou for your support and encourage. You have given me a chance to collabo-rate with excellent researchers, and exchange knowledge from multidiscipli-nary views. Per, from you, I have learned knowledge in mechanics, skills incoding, and most importantly, independent and critical thinking. Thank youboth for your patient guidance throughout the last four years.
I would like to thank my colleagues in the BMS, MIM and Engineering groups who have helped me. Thank you, Susanne, for sharing opinions in study and life, reminding me of fika as well as helping my Swedish sum-mary. Thanks Caroline for sharing your extensive experience in micro-CT and bone, and always taking time to answer my questions. Thanks Michael and Anna, for the wonderful discussion and help with my experiments. Thanks Jun and David for your help with bone cements. Thanks Fengzhen, Shaohui and Bengt for the discussion in mechanics. Thanks Alejandro and Joakim for your technical support in experimental setups. Thanks Jiaojiao and Rui for your help in porosity test. Thanks Xingxing for welding my electronic boards. Thanks Susan for checking my English writing. Thanks Wei, Le, Charlotte, Luimar, Gry, Céline, Lee, Camilla, Amina, Torbjörn, Marjamy, Yang and everyone in the group for making the working environment so lovely.
A special thank you to my co-authors and collaborators: Stephen Fergu-son for providing human bone, insights and valuable comments in my pro-jects and writing; Thomas Joffre for your support in in-situ experiments and DVC; Nico van Dijk for your amazing work and deep thoughts in DVC; Alicja Bojan and Philip Procter for your endless enthusiasm in the adhesive and screws that made the ex-situ pullout test possible; Anders Palmquist for sharing knowledge in micro-CT and helping me out with morphometric analysis; Andrea Spanou for your hard work in 3D-printing. Thanks to the Paul Scherrer Institute for the provision of synchrotron beamtime, and the beamline scientists, Anne Bonnin, Alessandra Patera, Federica Marone and Gordan Mikuljan, for their valuable help throughout the experiments. A heartful thank you to the ‘synchrotron squad’: Per, Cecilia, Caroline, Thomas, Sara G., Michael, Jenny, Susanne, Anna A.. Thank you for supporting each other and being problem solvers in challenging situations.
55
I also want to thank many friends I have made at Uppsala for making life much easier. Bengt and Pranee for introducing the city to me and your kind-hearted suggestions in every aspect. Ivón, Elias, Jinxing&Yingying, Alexey, Juan José, Gabriella, for the memorable afterwork hours. Yuanyuan, my roommate, and Zhen, thank you for your accompany during the years, good and hard times we have spent together, as well as your help with my study. Jun, again, for sharing life experience and nice food with me. The interesting talks and happy times we have during the numerous meals with your guys: Liyang, Bo Cao, Ling, Zhicheng, Dou, Tianbo, etc.
Last but not least, my sincere gratitude to my parents and brother.
56
References
[1] T.I.O. Foundation, Broken bones, broken lives: A roadmap to solve thefragility fracture crisis in Europe. (http://share.iofbonehealth.org/EU-6-Material/Reports /IOF%20Report_EU.pdf.)
[2] L.J. Melton III, E.J. Atkinson, M.K. O'Connor, W.M. O'Fallon, B.L. Riggs,Bone density and fracture risk in men, Journal of Bone and Mineral Research13(12) (1998) 1915-1923.
[3] A. Oden, E.V. McCloskey, J.A. Kanis, N.C. Harvey, H. Johansson, Burden ofhigh fracture probability worldwide: secular increases 2010-2040, OsteoporosisInternational 26(9) (2015) 2243-8.
[4] J.D. Barr, M.S. Barr, T.J. Lemley, R.M. McCann, Percutaneous vertebroplastyfor pain relief and spinal stabilization, Spine 25(8) (2000) 923-928.
[5] A.A. Uppin, J.A. Hirsch, L.V. Centenera, B.A. Pfiefer, A.G. Pazianos, I.S.Choi, Occurrence of new vertebral body fracture after percutaneousvertebroplasty in patients with osteoporosis, Radiology 226(1) (2003) 119-124.
[6] A.T. Trout, D.F. Kallmes, T.J. Kaufmann, New fractures after vertebroplasty:adjacent fractures occur significantly sooner, American Journal ofNeuroradiology 27(1) (2006) 217-223.
[7] G. Baroud, M. Bohner, Biomechanical impact of vertebroplasty, Joint BoneSpine 73(2) (2006) 144-150.
[8] G. Baroud, P. Heini, J. Nemes, M. Bohner, S. Ferguson, T. Steffen,Biomechanical explanation of adjacent fractures following vertebroplasty[letter], Radiology 229(2) (2003) 606-608.
[9] E. Lucchinetti, D. Thomann, G. Danuser, Review micromechanical testing ofbone trabeculae - potentials and limitations, Journal of Materials Science35(24) (2000) 6057-6065.
[10] H. Giambini, H.-J. Wang, C. Zhao, Q. Chen, A. Nassr, K.-N. An, Anterior andposterior variations in mechanical properties of human vertebrae measured bynanoindentation, Journal of Biomechanics 46(3) (2013) 456-461.
[11] B. Jost, C. Spross, H. Grehn, C. Gerber, Locking plate fixation of fractures ofthe proximal humerus: analysis of complications, revision strategies andoutcome, Journal of Shoulder and Elbow Surgery 22(4) (2013) 542-549.
[12] J.N. Weinstein, K.F. Spratt, D. Spengler, C. Brick, S. Reid, Spinal pediclefixation: reliability and validity of roentgenogram-based assessment andsurgical factors on successful screw placement, Spine 13(9) (1988) 1012-8.
[13] B.M. Frankel, T. Jones, C. Wang, Segmental polymethylmethacrylate-augmented pedicle screw fixation in patients with bone softening caused byosteoporosis and metastatic tumor involvement: a clinical evaluation,Neurosurgery 61(3) (2007) 531-538.
57
[14] G. Baroud, J. Nemes, P. Heini, T. Steffen, Load shift of the intervertebral discafter a vertebroplasty: a finite-element study, European Spine Journal 12(4)(2003) 421-426.
[15] A. El Saman, S. Meier, A. Sander, A. Kelm, I. Marzi, H. Laurer, Reducedloosening rate and loss of correction following posterior stabilization with orwithout PMMA augmentation of pedicle screws in vertebral fractures in theelderly, European Journal of Trauma and Emergency Surgery 39(5) (2013)455-460.
[16] D.H. Pahr, E. Dall'Ara, P. Varga, P.K. Zysset, HR-pQCT-based homogenisedfinite element models provide quantitative predictions of experimentalvertebral body stiffness and strength with the same accuracy as μFE models,Computer Methods in Biomechanics and Biomedical Engineering 15(7) (2012)711-720.
[17] C.J. Brown, R.A. Sinclair, A. Day, B. Hess, P. Procter, An approximate modelfor cancellous bone screw fixation, Computer Methods in Biomechanics andBiomedical Engineering 16(4) (2013) 443-450.
[18] A.J. Wirth, J. Goldhahn, C. Flaig, P. Arbenz, R. Müller, G.H. van Lenthe,Implant stability is affected by local bone microstructural quality, Bone 49(3)(2011) 473-478.
[19] Y. Chevalier, M. Matsuura, S. Krüger, C. Fleege, M. Rickert, M. Rauschmann,C. Schilling, Micro-CT and micro-FE analysis of pedicle screw fixation underdifferent loading conditions, Journal of Biomechanics 70 (2018) 204-211.
[20] S.K. Parashar, J.K. Sharma, A review on application of finite elementmodelling in bone biomechanics, Perspectives in Science 8 (2016) 696-698.
[21] R. Huiskes, E.Y.S. Chao, A survey of finite element analysis in orthopedicbiomechanics: The first decade, Journal of Biomechanics 16(6) (1983) 385-409.
[22] D.B. Burr, Chapter 1 - Bone morphology and organization, in: D.B. Burr, M.R.Allen (Eds.), Basic and Applied Bone Biology (Second Edition), AcademicPress2019, pp. 3-26.
[23] S.C. Cowin, S.B. Doty, Bone Tissue, in: S.C. Cowin, S.B. Doty (Eds.), TissueMechanics, Springer New York, New York, NY, 2007, pp. 341-384.
[24] T.M. Keaveny, E.F. Morgan, O.C. Yeh, Bone mechanics, Standard handbookof biomedical engineering and design, McGraw-Hill, New York, 2004, pp. 1-24.
[25] H.D. Barth, M.E. Launey, A.A. MacDowell, J.W. Ager, R.O. Ritchie, On theeffect of X-ray irradiation on the deformation and fracture behavior of humancortical bone, Bone 46(6) (2010) 1475-1485.
[26] M. Frank, D. Marx, V. Nedelkovski, J.-T. Fischer, D.H. Pahr, P.J. Thurner,Dehydration of individual bovine trabeculae causes transition from ductile toquasi-brittle failure mode, Journal of the Mechanical Behavior of BiomedicalMaterials 87 (2018) 296-305.
[27] J.-Y. Rho, G.M. Pharr, Effects of drying on the mechanical properties ofbovine femur measured by nanoindentation, Journal of Materials Science:Materials in Medicine 10(8) (1999) 485-488.
[28] N. Reznikov, M. Bilton, L. Lari, M.M. Stevens, R. Kröger, Fractal-likehierarchical organization of bone begins at the nanoscale, Science 360(6388)(2018) eaao2189.
58
[29] N. Reznikov, R. Shahar, S. Weiner, Bone hierarchical structure in threedimensions, Acta Biomaterialia 10(9) (2014) 3815-3826.
[30] G. Maquer, S.N. Musy, J. Wandel, T. Gross, P.K. Zysset, Bone volumefraction and fabric anisotropy are better determinants of trabecular bonestiffness than other morphological variables, Journal of Bone and MineralResearch 30(6) (2015) 1000-1008.
[31] S. Tassani, C. Öhman, M. Baleani, F. Baruffaldi, M. Viceconti, Anisotropy andinhomogeneity of the trabecular structure can describe the mechanical strengthof osteoarthritic cancellous bone, Journal of Biomechanics 43(6) (2010) 1160-1166.
[32] S.N. Musy, G. Maquer, J. Panyasantisuk, J. Wandel, P.K. Zysset, Not onlystiffness, but also yield strength of the trabecular structure determined by non-linear µFE is best predicted by bone volume fraction and fabric tensor, Journalof the Mechanical Behavior of Biomedical Materials 65 (2017) 808-813.
[33] D. Wu, P. Isaksson, S.J. Ferguson, C. Persson, Young’s modulus of trabecularbone at the tissue level: A review, Acta Biomaterialia 78 (2018) 1-12.
[34] D.L. Kopperdahl, T.M. Keaveny, Yield strain behavior of trabecular bone,Journal of Biomechanics 31(7) (1998) 601-608.
[35] E. Perilli, M. Baleani, C. Öhman, R. Fognani, F. Baruffaldi, M. Viceconti,Dependence of mechanical compressive strength on local variations inmicroarchitecture in cancellous bone of proximal human femur, Journal ofBiomechanics 41(2) (2008) 438-446.
[36] G. Bevill, F. Farhamand, T.M. Keaveny, Heterogeneity of yield strain in low-density versus high-density human trabecular bone, Journal of Biomechanics42(13) (2009) 2165-70.
[37] A. Rohlmann, H. Zilch, G. Bergmann, R. Kolbel, Material properties offemoral cancellous bone in axial loading, Archives of Orthopaedic andTraumatic Surgery 97(2) (1980) 95-102.
[38] R. Carretta, E. Stüssi, R. Müller, S. Lorenzetti, Within subject heterogeneity intissue-level post-yield mechanical and material properties in human trabecularbone, Journal of the Mechanical Behavior of Biomedical Materials 24 (2013)64-73.
[39] B. Busse, M. Hahn, M. Soltau, J. Zustin, K. Püschel, G.N. Duda, M. Amling,Increased calcium content and inhomogeneity of mineralization render bonetoughness in osteoporosis: Mineralization, morphology and biomechanics ofhuman single trabeculae, Bone 45(6) (2009) 1034-1043.
[40] M.E. Szabó, M. Taylor, P.J. Thurner, Mechanical properties of single bovinetrabeculae are unaffected by strain rate, Journal of Biomechanics 44(5) (2011)962-967.
[41] E.W. Andrews, G. Gioux, P. Onck, L.J. Gibson, Size effects in ductile cellularsolids. Part II: experimental results, Int J Mech Sci 43(3) (2001) 701-713.
[42] A.J. Wirth, R. Müller, G. Harry van Lenthe, The discrete nature of trabecularbone microarchitecture affects implant stability, Journal of Biomechanics 45(6)(2012) 1060-1067.
[43] R. Zdero, M.S.R. Aziz, B. Nicayenzi, Chapter 8 - Pullout force testing ofcortical and cancellous screws in whole bone, in: R. Zdero (Ed.), ExperimentalMethods in Orthopaedic Biomechanics, Academic Press2017, pp. 117-132.
59
[44] B.D. Elder, S.-F.L. Lo, C. Holmes, C.R. Goodwin, T.A. Kosztowski, I.A. Lina,J.E. Locke, T.F. Witham, The biomechanics of pedicle screw augmentationwith cement, The Spine Journal 15(6) (2015) 1432-1445.
[45] S. Larsson, T.W. Bauer, Use of injectable calcium phosphate cement forfracture fixation: A review, Clinical Orthopaedics and Related Research 395(2002) 23-32.
[46] F. Lattig, Bone cement augmentation in the prevention of adjacent segmentfailure after multilevel adult deformity fusion, Journal of Spinal Disorders &Techniques 22(6) (2009) 439-43.
[47] D.C. Moore, R.S. Maitra, L.A. Farjo, G.P. Graziano, S.A. Goldstein,Restoration of pedicle screw fixation with an in situ setting calcium phosphatecement, Spine 22(15) (1997) 1696-705.
[48] P.E. Paré, J.L. Chappuis, R. Rampersaud, A.O. Agarwala, J.H. Perra, S. Erkan,C. Wu, Biomechanical evaluation of a novel fenestrated pedicle screwaugmented with bone cement in osteoporotic spines, Spine 36(18) (2011)E1210-E1214.
[49] S.M. Renner, T.H. Lim, W.J. Kim, L. Katolik, H.S. An, G.B. Andersson,Augmentation of pedicle screw fixation strength using an injectable calciumphosphate cement as a function of injection timing and method, Spine 29(11)(2004) E212-6.
[50] K.S. Leung, W.S. Siu, S.F. Li, L. Qin, W.H. Cheung, K.F. Tam, P.P.Y. Lui, Anin vitro optimized injectable calcium phosphate cement for augmenting screwfixation in osteopenic goats, Journal of Biomedical Materials Research Part B:Applied Biomaterials 78B(1) (2006) 153-160.
[51] S. Larsson, V.A. Stadelmann, J. Arnoldi, M. Behrens, B. Hess, P. Procter, M.Murphy, D.P. Pioletti, Injectable calcium phosphate cement for augmentationaround cancellous bone screws. In vivo biomechanical studies, Journal ofBiomechanics 45(7) (2012) 1156-60.
[52] B.E. McKoy, Y.H. An, An injectable cementing screw for fixation inosteoporotic bone, Journal of Biomedical Materials Research 53(3) (2000) 216-20.
[53] B. Bai, F.J. Kummer, J. Spivak, Augmentation of anterior vertebral body screwfixation by an injectable, biodegradable calcium phosphate bone substitute,Spine 26(24) (2001) 2679-83.
[54] B.-S. Yu, Z.-M. Li, Z.-Y. Zhou, L.-W. Zeng, L.-B. Wang, Z.-M. Zheng, W.W.Lu, Biomechanical effects of insertion location and bone cement augmentationon the anchoring strength of iliac screw, Clinical Biomechanics 26(6) (2011)556-561.
[55] D. Liu, Z.-x. Wu, X.-m. Pan, S.-c. Fu, M.-x. Gao, L. Shi, W. Lei,Biomechanical comparison of different techniques in primary spinal surgery inosteoporotic cadaveric lumbar vertebrae: expansive pedicle screw versuspolymethylmethacrylate-augmented pedicle screw, Archives of Orthopaedicand Trauma Surgery 131(9) (2011) 1227-1232.
[56] J.S. Sarzier, A.J. Evans, D.W. Cahill, Increased pedicle screw pullout strengthwith vertebroplasty augmentation in osteoporotic spines, Journal ofNeurosurgery: Spine 96(3) (2002) 309-312.
[57] P. Procter, P. Bennani, C.J. Brown, J. Arnoldi, D.P. Pioletti, S. Larsson,Variability of the pullout strength of cancellous bone screws with cementaugmentation, Clinical Biomechanics 30(5) (2015) 500-506.
60
[58] H.J. Seeherman, M. Bouxsein, H. Kim, R. Li, X.J. Li, M. Aiolova, J.M.Wozney, Recombinant human bone morphogenetic protein-2 delivered in aninjectable calcium phosphate paste accelerates osteotomy-site healing in anonhuman primate model, Journal of Bone and Joint Surgery. AmericanVolume 86-a(9) (2004) 1961-1972.
[59] V.A. Stadelmann, E. Bretton, A. Terrier, P. Procter, D.P. Pioletti, Calciumphosphate cement augmentation of cancellous bone screws can compensate forthe absence of cortical fixation, Journal of Biomechanics 43(15) (2010) 2869-2874.
[60] M. Pujari-Palmer, H. Guo, D. Wenner, H. Autefage, C.D. Spicer, M.M.Stevens, O. Omar, P. Thomsen, M. Edén, G. Insley, P. Procter, H. Engqvist, Anovel class of injectable bioceramics that glue tissues and biomaterials,Materials 11(12) (2018) 2492.
[61] R. Jungmann, M.E. Szabó, G. Schitter, R. Yue-Sing Tang, D. Vashishth, P.K.Hansma, P.J. Thurner, Local strain and damage mapping in single trabeculaeduring three-point bending tests, Journal of the Mechanical Behavior ofBiomedical Materials 4(4) (2011) 523-534.
[62] P.J. Thurner, B. Erickson, R. Jungmann, Z. Schriock, J.C. Weaver, G.E.Fantner, G. Schitter, D.E. Morse, P.K. Hansma, High-speed photography ofcompressed human trabecular bone correlates whitening to microscopicdamage, Engineering Fracture Mechanics 74(12) (2007) 1928-1941.
[63] P.J. Thurner, P. Wyss, R. Voide, M. Stauber, M. Stampanoni, U. Sennhauser,R. Müller, Time-lapsed investigation of three-dimensional failure and damageaccumulation in trabecular bone using synchrotron light, Bone 39(2) (2006)289-299.
[64] M. Palanca, G. Tozzi, L. Cristofolini, M. Viceconti, E. Dall'Ara, Three-dimensional local measurements of bone strain and displacement: comparisonof three digital volume correlation approaches, Journal of Biomechanicalengineering 137(7) (2015) 071006-071006.
[65] G. Tozzi, E. Dall’Ara, M. Palanca, M. Curto, F. Innocente, L. Cristofolini,Strain uncertainties from two digital volume correlation approaches inprophylactically augmented vertebrae: Local analysis on bone and cement-bone microstructures, Journal of the Mechanical Behavior of BiomedicalMaterials 67(Supplement C) (2017) 117-126.
[66] S. Yamada, S. Tadano, K. Fukasawa, Micro-cantilever bending for elasticmodulus measurements of a single trabecula in cancellous bone, Journal ofBiomechanics 49(16) (2016) 4124-4127.
[67] P. Lucksanasombool, W.A. Higgs, M. Ignat, R.J. Higgs, M.V. Swain,Comparison of failure characteristics of a range of cancellous bone-bonecement composites, Journal of biomedical materials research. Part A 64(1)(2003) 93-104.
[68] T.L. Mueller, S.E. Basler, R. Müller, G.H. van Lenthe, Time-lapsed imaging ofimplant fixation failure in human femoral heads, Medical Engineering &Physics 35(5) (2013) 636-643.
[69] F. Gillard, R. Boardman, M. Mavrogordato, D. Hollis, I. Sinclair, F. Pierron,M. Browne, The application of digital volume correlation (DVC) to study themicrostructural behaviour of trabecular bone during compression, Journal ofthe Mechanical Behavior of Biomedical Materials 29 (2014) 480-499.
61
[70] M.K. Ryan, A.A. Mohtar, T.M. Cleek, K.J. Reynolds, Time-elapsed screwinsertion with microCT imaging, Journal of Biomechanics 49(2) (2016) 295-301.
[71] T. Joffre, P. Isaksson, P. Procter, C. Persson, Trabecular deformations duringscrew pull-out: a micro-CT study of lapine bone, Biomechanics and Modelingin Mechanobiology 16(4) (2017) 1349-1359.
[72] International standard ISO 5833 Implants for surgery—Acrylic resin cements,International Organization for Standardization Geneva, (2002).
[73] ASTM Standard F543-07: Standard specification and test methods for metallicmedical bone screws, PA, USA: ASTM International, West Conshohocken(2007).
[74] M. Dierolf, A. Menzel, P. Thibault, P. Schneider, C.M. Kewish, R. Wepf, O.Bunk, F. Pfeiffer, Ptychographic X-ray computed tomography at the nanoscale,Nature 467 (2010) 436.
[75] M. Peña Fernández, S. Cipiccia, E. Dall’Ara, A.J. Bodey, R. Parwani, M. Pani,G.W. Blunn, A.H. Barber, G. Tozzi, Effect of SR-microCT radiation on themechanical integrity of trabecular bone using in situ mechanical testing anddigital volume correlation, Journal of the Mechanical Behavior of BiomedicalMaterials 88 (2018) 109-119.
[76] C.J. Hernandez, D.S. Ramsey, S.J. Dux, E.H. Chu, C.M. Rimnac, Irradiationdoes not modify mechanical properties of cancellous bone under compression,Clinical Orthopaedics and Related Research 470(9) (2012) 2488-2495.
[77] S.J. Dux, D. Ramsey, E.H. Chu, C.M. Rimnac, C.J. Hernandez, Alterations indamage processes in dense cancellous bone following gamma-radiationsterilization, Journal of Biomechanics 43(8) (2010) 1509-1513.
[78] M. Stampanoni, R. Abela, G. Borchert, B.D. Patterson, New developments insynchrotron-based microtomography, SPIE2004 169-181.
[79] A.M. Parfitt, M.K. Drezner, F.H. Glorieux, J.A. Kanis, H. Malluche, P.J.Meunier, S.M. Ott, R.R. Recker, Bone histomorphometry: standardization ofnomenclature, symbols, and units. Report of the ASBMR HistomorphometryNomenclature Committee, Journal of Bone and Mineral Research 2(6) (1987)595-610.
[80] T. Hildebrand, P. Rüegsegger, A new method for the model-independentassessment of thickness in three-dimensional images, Journal of Microscopy185(1) (1997) 67-75.
[81] F. Forsberg, M. Sjödahl, R. Mooser, E. Hack, P. Wyss, Full three-dimensionalstrain measurements on wood exposed to three-point bending: Analysis by useof digital volume correlation applied to synchrotron radiation micro-computedtomography image data, Strain 46(1) (2010) 47-60.
[82] C. Chateau, T.T. Nguyen, M. Bornert, J. Yvonnet, DVC-based imagesubtraction to detect microcracking in lightweight concrete, Strain 54(5) (2018)e12276.
[83] Z. Yang, W. Ren, R. Sharma, S. McDonald, M. Mostafavi, Y. Vertyagina, T.J.Marrow, In-situ X-ray computed tomography characterisation of 3D fractureevolution and image-based numerical homogenisation of concrete, Cement andConcrete Composites 75 (2017) 74-83.
[84] C.M. Disney, A. Eckersley, J.C. McConnell, H. Geng, A.J. Bodey, J.A.Hoyland, P.D. Lee, M.J. Sherratt, B.K. Bay, Synchrotron tomography ofintervertebral disc deformation quantified by digital volume correlation reveals
62
microstructural influence on strain patterns, Acta Biomaterialia 92 (2019) 290-304.
[85] C. Jailin, A. Bouterf, M. Poncelet, S. Roux, In situ μ CT-scan mechanical tests:Fast 4D mechanical identification, Experimental Mechanics 57(8) (2017)1327-1340.
[86] B.K. Bay, T.S. Smith, D.P. Fyhrie, M. Saad, Digital volume correlation: Three-dimensional strain mapping using X-ray tomography, Experimental Mechanics39(3) (1999) 217-226.
[87] M. Shakoor, A. Buljac, J. Neggers, F. Hild, T.F. Morgeneyer, L. Helfen, M.Bernacki, P.-O. Bouchard, On the choice of boundary conditions formicromechanical simulations based on 3D imaging, International Journal ofSolids and Structures 112(Supplement C) (2017) 83-96.
[88] B.C. Roberts, E. Perilli, K.J. Reynolds, Application of the digital volumecorrelation technique for the measurement of displacement and strain fields inbone: A literature review, Journal of Biomechanics 47(5) (2014) 923-934.
[89] E. Bar-Kochba, J. Toyjanova, E. Andrews, K.-S. Kim, C. Franck, A fastiterative digital volume correlation algorithm for large deformations,Experimental Mechanics 55(1) (2015) 261-274.
[90] F. Forsberg, X-ray microtomography and digital volume correlation forinternal deformation and strain analysis, Luleå University of Technology,Luleå tekniska universitet, Luleå, 2008.
[91] B. van Rietbergen, H. Weinans, R. Huiskes, A. Odgaard, A new method todetermine trabecular bone elastic properties and loading usingmicromechanical finite-element models, Journal of Biomechanics 28(1) (1995)69-81.
[92] N.P. van Dijk, D. Wu, C. Persson, P. Isaksson, A global digital volumecorrelation algorithm based on higher-order finite elements: Implementationand evaluation, International Journal of Solids and Structures 168 (2019) 211-227.
[93] S.V.N. Jaecques, H. Van Oosterwyck, L. Muraru, T. Van Cleynenbreugel, E.De Smet, M. Wevers, I. Naert, J. Vander Sloten, Individualised, micro CT-based finite element modelling as a tool for biomechanical analysis related totissue engineering of bone, Biomaterials 25(9) (2004) 1683-1696.
[94] G. Niebur, T. Keaveny, Computational Modeling of Trabecular BoneMechanics, in: S. De, F. Guilak, M. Mofrad R. K (Eds.), ComputationalModeling in Biomechanics, Springer Netherlands2010, pp. 277-306.
[95] E. Verhulp, B. van Rietbergen, R. Müller, R. Huiskes, Indirect determinationof trabecular bone effective tissue failure properties using micro-finite elementsimulations, Journal of Biomechanics 41(7) (2008) 1479-1485.
[96] C.J. Stankewich, M.F. Swiontkowski, A.F. Tencer, D.N. Yetkinler, R.D. Poser,Augmentation of femoral neck fracture fixation with an injectable calcium-phosphate bone mineral cement, Journal of Orthopaedic Research 14(5) (1996)786-793.
[97] A. Piper, C.J. Brown, A computational approximation to model variation incancellous bone screw pull-out, Journal of Medical Devices 10(2) (2016)021001-1-7.
[98] C.A. Reitman, L. Nguyen, G.R. Fogel, Biomechanical evaluation ofrelationship of screw pullout strength, insertional torque, and bone mineral
63
density in the cervical spine, Journal of spinal disorders & techniques 17(4)(2004) 306-311.
[99] K. Okuyama, E. Abe, T. Suzuki, Y. Tamura, M. Chiba, K. Sato, Influence ofbone mineral density on pedicle screw fixation: a study of pedicle screwfixation augmenting posterior lumbar interbody fusion in elderly patients, TheSpine Journal 1(6) (2001) 402-407.
[100]Y. Liu, J. Xu, D. Sun, F. Luo, Z. Zhang, F. Dai, Biomechanical and finiteelement analyses of bone cement-Injectable cannulated pedicle screw fixationin osteoporotic bone, Journal of Biomedical Materials Research Part B:Applied Biomaterials 104(5) (2016) 960-967.
[101]L.J. Gibson, M.F. Ashby, Cellular solids: Structure and properties, 2 ed.,Cambridge University Press, Cambridge, 1997.
Acta Universitatis UpsaliensisDigital Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Science and Technology 1836
Editor: The Dean of the Faculty of Science and Technology
A doctoral dissertation from the Faculty of Science andTechnology, Uppsala University, is usually a summary of anumber of papers. A few copies of the complete dissertationare kept at major Swedish research libraries, while thesummary alone is distributed internationally throughthe series Digital Comprehensive Summaries of UppsalaDissertations from the Faculty of Science and Technology.(Prior to January, 2005, the series was published under thetitle “Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Science and Technology”.)
Distribution: publications.uu.seurn:nbn:se:uu:diva-385143
ACTAUNIVERSITATIS
UPSALIENSISUPPSALA
2019