Post on 11-Jun-2020
In the format provided by the authors and unedited.
Precision assessment of label-free psoriasis biomarkers with ultra-broadband optoacoustic mesoscopy
Juan Aguirre1, Mathias Schwarz 1,Natalie Garzorz2, Murad Omar1, Andreas Buehler1, Kilian Eyerich2, Vasilis Ntziachristos1*
1. Chair for Biological Imaging, Technische Universität München and Institute for Biological and Medical Imaging, Helmholtz Zentrum München,IngolstädterLandstr.1, D85764 Neuherberg, Germany.
2. Division of Environmental Dermatology and Allergy, Helmholtz Center Munich/Technische Universität Munich and
ZAUM —Center for Allergy and Environment, Technische Universität Munich, Munich, Germany.
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
SUPPLEMENTARY INFORMATIONVOLUME: 1 | ARTICLE NUMBER: 0068
NATURE BIOMEDICAL ENGINEERING | DOI: 10.1038/s41551-017-0068 | www.nature.com/natbiomedeng 1
TABLE OF CONTENTS
Suppl Note 1: UB-RSOM setup, Frequency Equalization and interpretation of images.
Page 3
Suppl Note 2: Effect of bandwidth on optoacoustic mesoscopy of the skin.
Page 5
Suppl Note 3: Imaging of eczema, vasculitis, hair and nevus.
Page 7
Suppl Note 4: Characterization of the system.
Page 9
Suppl Note 5: Comparison of frequency band equalization with logarithmic compression. Page 12
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE BIOMEDICAL ENGINEERING | DOI: 10.1038/s41551-017-0068 | www.nature.com/natbiomedeng 2
SUPPLEMENTARY INFORMATION
Supplementary Note 1:
UB-RSOM setup, Frequency Equalization and interpretation of images.
Suppl Fig. 1a provides details of the RSOM imaging system by showing a scheme of its elements and their
connections and Suppl Fig. 1b demonstrates the dependence of frequency equalization on the parameter
(RSOM system components and the frequency equalization algorithm are analytically discussed in methods).
As evident on Suppl Fig. 1b, different a values result in images of different appearance, weighting differently
the spatial frequencies contained. For this reason, optimization of the parameter (see methods Eq.1,2) is
critical in order to achieve maximum imaging performance. Specifically, if the selected value of is low (for
example =8.2), the minimal value of the functional is not reached. In such case, the low frequency
components of a typical skin image predominate (red). As a consequence the epidermis cannot be
distinguished, from the dermis and the smallest vessels next to the epidermal-dermal junction cannot be
resolved from the junction itself. An analogous effect occurs if the value of is too high. In such case, the
high frequency components of the image predominate (green). As a consequence, some of the bigger vessels
in the lower plexus are not visible and it is not possible to differentiate the dermis from the epidermis.
In order to aid in the interpretation and comprehension of RSOM skin images, Suppl Fig. 1c show a
comparison of a cross sectional view of healthy skin together with an artistic representation of the epidermis
and the dermal skin microvasculature. The canonical structural components of the vessels structure (superficial
plexus, connecting vessels and lower plexus) can be easily identified in the RSOM image.
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE BIOMEDICAL ENGINEERING | DOI: 10.1038/s41551-017-0068 | www.nature.com/natbiomedeng 3
SUPPLEMENTARY INFORMATION
Supplementary figure 1: Experimental Setup and Frequency Equalization Performance. a) Scheme of the
acquisition setup. MS motorized stages, PD photodiode, TRD, transducer, INTC interface unit. The transducer is scanned in the x-y direction acquiring photoacoustic signals (A-lines). b) the graph shows the function obtained when the squared error is plotted as a function of the parameter α. By selecting the alpha value corresponding to the minimum of the curve (α =16.4) the FBC allows to differentiate some layers of the epidermis and the vascular structure thanks to the RGB representation. If the parameter correspond to region of the curve out of the minimum (for example α =8.2) or α =22, the contrast is lost. c) Comparison of the RSOM cross sectional image corresponding to the forearm of a healthy volunteer and an artistical representation of the canonical skin model. EP epidermis SP stands for superficial plexus, CV is connecting vessels and DP is the deep plexus. Scale bars: 200 μm
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE BIOMEDICAL ENGINEERING | DOI: 10.1038/s41551-017-0068 | www.nature.com/natbiomedeng 4
SUPPLEMENTARY INFORMATION
Supplementary Note 2:
Effect of bandwidth on optoacoustic mesoscopy of the skin
The dependence of the imaging performance achieved in optoacoustic mesoscopy when using
different bandwidths is shown in Suppl Fig. 2. We studied the appearance of imaged at four different
frequency bands: 10-40 MHz, 10-80 MHz, 10-120 MHz and 10-180 MHz.
Coronal views of the dermal vasculature do not show any significant difference in the ability to
resolve the vessel structure when comparing the 10-180 MHz and the 10-120 MHz band. Some blurring can be
appreciated in the dermal vessels seen on the 10-80 MHz band. Finally the 10-40 MHz band shows images
markedly affected by blurring (Suppl Fig. 2.b). When imaging the capillary loops, there is a clear difference
between the 10-180 MHz and the 10-120 MHz reconstructions (Suppl Fig. 2.c), which can be appreciated
when calculating the diameter of a representative capillary loops, showing an FWHM of 24 µm for the 10-180
MHz band over 34 µm for the 10-120 MHz band and 80 µm in the 10-80 MHz band. We note that these
capillaries cannot be visualized in the 10-40 MHz band (Suppl Fig. 2.d,e).
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE BIOMEDICAL ENGINEERING | DOI: 10.1038/s41551-017-0068 | www.nature.com/natbiomedeng 5
SUPPLEMENTARY INFORMATION
Supplementary figure 2 Effect of bandwidth on optoacoustic mesoscopy of skin a) Equalized RSOM reconstruction of healthy skin. b) Coronal views of the skin vasculature for reconstruction at different wavelengths. c) Coronal views of the capillary loops reconstructed at different frequency bands. d) Zoomed area corresponding to the zone pointed by the green arrows at the 10-180 frequency band and the 10-40 frequency band e) Measured diameter (lateral direction) of a representative capillary loop reconstructed at different frequency bands. Scale bars: 500 μm
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE BIOMEDICAL ENGINEERING | DOI: 10.1038/s41551-017-0068 | www.nature.com/natbiomedeng 6
SUPPLEMENTARY INFORMATION
Supplementary Note 3:
Imaging of eczema, vasculitis, hair and nevus
In addition to psoriasis, a number of additional skin features and conditions have been imaged. We
demonstrate here the appearance of eczema, which is associated with a strong inflammatory process. The cross
sections corresponding to skin areas affected by eczema of two different patients (Suppl Fig. 3a,b) exhibit
different degree of acanthosis and elongated capillary loops. Differences in the vascular structures in the
dermis can be clearly observed. In Suppl Fig.3c-g we show the quantification of different vascular and
epidermal properties corresponding to three eczema patients. In analogy to the analysis of psoriasis, the
features resolved included the total blood volume, epidermis thickness, fractal number, density of capillaries
and capillaries diameter. Such features can be used as precision biomarkers for disease severity evaluation and
can be particularly useful for allergy testing, being the latter a process driven by subjective visual
observations.
Suppl Fig. 3h shows a cross section of healthy skin in which a hair shaft can be observed together with its
surrounding vessels. The length of the hair inside the skin is ~2mm. The ability to image hair follicles and
their vascular environment may be employed in the study of alopecia and acne. The ability of the FBE method
to image nevi and its underlying vascular structure is showcased in Suppl Fig. 3i. The nevi can be clearly
observed in the epidermis. Eventual changes in the nevi shape and size or in the surrounding vessel structure
can be a sign of malignant transformation. Finally, the UB-RSOM cross section of a skin area affected by
vasculitis is shown in Suppl Fig. 3j. The vessel structure appears completely distorted as corresponds to the
clinical phenotype of the condition.
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE BIOMEDICAL ENGINEERING | DOI: 10.1038/s41551-017-0068 | www.nature.com/natbiomedeng 7
SUPPLEMENTARY INFORMATION
Supplementary figure 3 UB-RSOM imaging of eczema and nevus a,b) Cross sectional images of skin regions affected by eczema. The tip of capillary loops that climb towards the surface appear in the images (white arrows) c-g) Measured values related to the vascular structure and epidermis thickness of three different eczema patients including the ones corresponding to figure a and b. h) Cross sectional image of healthy skin in which a hair shaft is displayed (white arrows) . i) Cross sectional image of a healthy skin region. A benign nevus is displayed (white arrows). j) Cross sectional image of a skin area affected by vasculitis. The vascular structure is completely distorted compared to the appearance of healthy skin. Scale bar: 200 μm.
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE BIOMEDICAL ENGINEERING | DOI: 10.1038/s41551-017-0068 | www.nature.com/natbiomedeng 8
SUPPLEMENTARY INFORMATION
Supplementary Note 4:
Characterization of the system
We have characterized the lateral resolution (LR) and axial resolution (AR) of the system for the 10-180 MHz
and the 10-120 MHz band implementation by imaging a 3.5 µm diameter carbon microsphere inserted in an
agar based gelatin phantom at a depth of 300 µm. India ink and intralilpid was added to the phantom in order
to mimic the optical properties of tissue. We defined the LR taken as the FWHM of a profile of the
microsphere in the lateral direction and the AR as the FWHM of a profile taken in the axial direction (Suppl
Fig. 4a,b). For the 10-180 MHz implementation we obtained a LR of 18.4 µm and a AR 4.5 µm. For the 10-
120 MHz implementation we obtained a LR of 29.6 µm and a AR 8.6 µm.
In order to characterize the possible degradation of the resolution as a function of depth we have imaged a
USP 11-0 surgical suture (diameter ~10-19 µm) immersed in a phantom made from agar gelatin at different
depths ranging from 350 µm to 2500 µm (~350 µm/upperdermis, ~800 µm/dermis and ~1500 µm/fat layer) .
Intralipid and ink were added to the gel in order to mimic the optical properties of tissue. At each depth the
profile of the suture was obtained in the lateral direction for the 10-180 MHz and the 10-120 MHz
implementation. The FWHM of the profiles in the lateral direction at each depth were calculated (Suppl Fig.
4c). The degradation of the resolution at a certain depth can be characterized as difference in the FWHM
obtained for the suture at such depth and the FWHM of the most superficial suture. While the resolution
remained invariable through the whole depth of the dermis, certain degradation (~5 µm ) could be observed at
the depths corresponding to the fat layer .
We have also characterized the sensitivity of the system when a vessel conforms a certain angle with the axial
axis of the transducer (Suppl Fig. 4d). To do so we have built a phantom made out of agar and we have
introduced a USP 11-0 surgical suture on it. The suture was rotated confirming different angles with regard to
the axial axis of the transducer. For each angle the sensitivity (sen) of the system was estimated (Suppl Fig.
4d-e) using the flowing formula: (1) where are the voxel values that fulfill the following
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE BIOMEDICAL ENGINEERING | DOI: 10.1038/s41551-017-0068 | www.nature.com/natbiomedeng 9
SUPPLEMENTARY INFORMATION
expression: Such operation was done for the 10-180 MHz implementation and for 10-
120 MHz implementation. From the suture reconstructed at different angles, the diameter was calculated as the
FWHM of a profile in thedirection of the axial axis of the suture (Suppl Fig. 4d) in the 10-120 MHz
implementation. The maximum discrepancy was of ~4µm.
It is important to note that acoustic absorption behaves slightly different in water than in tissue. The acoustic
attenuation coefficient follows the empirical law: where is sound frequency, is the acoustic
attenuation constant and is the exponent of the power law. Typically DB/MHz2cm and in
tissue and DB/MHz2cm and in water. At high frequencies (~200 MHz) the acoustic
absorption of water matches the acoustic absorption of tissue, whereas at low frequencies tissue absorbs sound
better than water (Suppl Fig. 4g).
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE BIOMEDICAL ENGINEERING | DOI: 10.1038/s41551-017-0068 | www.nature.com/natbiomedeng 10
SUPPLEMENTARY INFORMATION
Supplementary figure 4: a)axial profile of a 3.5 µm diameter microsphere inserted in a tissue phantom at a depth of 300 µm obtained with the 10-180MHz and the 10-120 MHz implementation. b) lateral profile of a 3.5 µm diameter microsphere inserted in a tissue phantom at a depth of 300 µm obtained with the 10-180MHz and the 10-120 MHz implementation. c)FWHM obtained from a profile of a USP 11-0 suture taken in the lateral direction . The suture was placed at different depths in a tissue phantom. UDR corresponds to upper dermis, DR corresponds to dermis and FL corresponds to the fat layer. The FWHM maximum is constant through the whole skin depth d) scheme of the experimental set-up used to characterize the system performance when a vessel conforms a certain angle with the axial axis of the transducer. sut stands for suture, tx stands for the transducer axial axis and sx defines the suture axial axis. e) sensitivity of the system as function of the angle β as depicted in (d). f) FWHM of the suture in the axial direction obtained by the system in the 10-120 MHz implementation as function of the angle β as depicted in (d). g) Acoustic attenuation coefficient as function of frequency for sound propagating in tissue and in water.
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE BIOMEDICAL ENGINEERING | DOI: 10.1038/s41551-017-0068 | www.nature.com/natbiomedeng 11
SUPPLEMENTARY INFORMATION
Supplementary Note 5:
Comparison of frequency band equalization with logarithmic compression.
We have compared the FBE method with image enhancement based on logarithmic compression, which is
often used in ultrasound imaging. Logarithmic compression of a signal (or image) on vector is calculated
using the formula: where is the i-th element of and a is the compression
factor. For our comparison, we acquired data from healthy skin covering an area of 4 x 2 mm (266 x 135 scan
points) using the 10-120 MHz implementation of UB-RSOM. Suppl Fig. 5a shows the image reconstructed
using the entire frequency range, without any kind of equalization. Suppl Fig. 5b shows improved contrast
and sharper features after applying logarithmic compression. Suppl Fig. 5c shows the comparably good
contrast and sharp image features after reconstructing the image using the FBE method. Epidermal structures
are more easily distinguished from the upper vascular plexus in the FBE image (Suppl Fig. 5d-f). Suppl Fig.
5g shows that small vessels (10-20 μm) lying deeper in the tissue and emitting predominantly high-frequency
optoacoustic signals are resolved better following the FBE method. In these comparisons, logarithmic
compression was applied using a values ranging from 10-6 to 103, and only the best results are shown here.
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE BIOMEDICAL ENGINEERING | DOI: 10.1038/s41551-017-0068 | www.nature.com/natbiomedeng 12
SUPPLEMENTARY INFORMATION
Supplementary figure 5:(a) Cross-sectional image of healthy skin reconstructed with no frequency band equalization. (b) Cross-sectional image of the same region reconstructed without frequency band equalization and with logarithmic compression. (c) Cross-sectional image of the same region reconstructed with frequency band equalization. (d-f) Zoomed-in regions corresponding to the upper dashed squares in panels (a)-(c). Frequency band equalization allows the epidermis (EP) in panel (f) to be distinguished from the underlying microvascular plexus (arrows). (g) Optoacoustic signal along the microvessel cross-section indicated by a colored dashed line in panels (a)-(c). While log compression provides similar results than no-equalization reconstruction, the FBE method provides the sharpest representation of the vessel. REFERENCES 1 Omar, M., Schwarz, M., Soliman, D., Symvoulidis, P. & Ntziachristos, V. Optical imaging of post-embryonic zebrafish using multi orientation raster scan optoacoustic mesoscopy (RSOM). Light: Science & Applications (2017) 6, e16186; doi:10.1038/lsa.2016.186
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE BIOMEDICAL ENGINEERING | DOI: 10.1038/s41551-017-0068 | www.nature.com/natbiomedeng 13
SUPPLEMENTARY INFORMATION