Innovative Food Science and Emerging Technologies 22 (2014) 137–146

Contents lists available at ScienceDirect

Innovative Food Science and Emerging Technologies

j ourna l homepage: www.e lsev ie r .com/ locate / i fset

Understanding near infrared radiation propagation in pig skinreflectance measurements

Eduardo Zamora-Rojas a,⁎, Ana Garrido-Varo a, Ben Aernouts b, Dolores Pérez-Marín a, Wouter Saeys b,Yukio Yamada c, José Emilio Guerrero-Ginel a

a Department of Animal Production, Faculty of Agricultural and Forestry Engineering, University of Córdoba, Agrifood Campus of International Excellence (ceiA3), Campus Rabanales,N-IV, Km 396, Córdoba 14014, Spainb Department of Biosystems, Division of Mechatronics, Biostatistics and Sensors, University of Leuven, Kasteelpark Arenberg 30, Leuven 3001, Belgiumc Department of Mechanical Engineering and Intelligent Systems, University of Electro-Communications, Chofu, Tokyo 182-8585, Japan

⁎ Corresponding author.E-mail address: e.zamora@uco.es (E. Zamora-Rojas).

1466-8564/$ – see front matter © 2014 Elsevier Ltd. All rihttp://dx.doi.org/10.1016/j.ifset.2014.01.006

a b s t r a c t

a r t i c l e i n f o

Article history:Received 25 August 2013Accepted 18 January 2014Available online 26 January 2014

Editor Proof Receive Date 13 February 2014

Keywords:Monte CarloSimulationNear Infrared Reflectance SpectroscopyLight propagationSpatially resolved spectroscopySkinPig

Non-invasive and non-destructive analysis based on Near Infrared Reflectance Spectroscopy (NIRS) sensors areused for the development of quality control systems in numerous in situ/on-line applications in different fields.However, few studies have been published investigating and trying to understand how light propagates throughagri-food tissues. In this paper, diffuse reflectance spectra of Iberian pig skinwere simulated for different source–detector distances in the wavelength range from 1150 nm to 1850 nm. The average photon visit depth and thefraction of absorbed energy indicated that most of the light was absorbed by the dermis layer. Nevertheless, larg-er source–detector distances enabled to acquire information from deeper in the skin tissue and to maximize thesensitivity of the captured signal to the subcutaneous adipose tissue. These simulation results were experimen-tally validated using spatially resolved reflectance spectroscopy, and a good agreement was obtained betweenthe simulations and measurements.Industrial relevance: A stochastic approach such asMonte Carlomethod is used for investigating and characteriz-ing non-destructive and non-invasive diffuse reflectance NIR spectroscopy measurements of Iberian pigs in the1150–1850 nm wavelength range. Different source detector distances have been evaluated in order to optimizethe optical configuration of NIRS instruments andmaximize the spectral information detected from the subcuta-neous adipose tissue. This study helps to understand how the light is propagated in pig tissues, which is relevantto design new spectrometers that can be used as non-destructive in situ/in vivo quality control systemor support-decision making system.

© 2014 Elsevier Ltd. All rights reserved.

1. Introduction

Near Infrared Reflectance Spectroscopy (NIRS) has been investigat-ed for multiple non-invasive in vivo measurements, especially in medi-cal applications (Batten, 2012). NIR radiation can penetrate into tissuesoffering a potential spectral windowwithout the hazards of ionizing ra-diation. This raised the interest of the agri-food sector, which is movingfromdestructive at-line analysis, with extensive sample presentation, toin situ/on-line/in-field applications on intact products.

In the meat industry, for instance, NIR spectroscopic measurementsacquired through the skin of animals would enable fast and sustainablecollection of information on the subcutaneous tissues than other analy-sis modes. Moreover, in the pig sector, in vivo NIRS analysis can be avaluable alternative to the banned practice of subcutaneous adipose tis-sue biopsies to optimize the complementary feeding at the final rearingperiod in order to assure a homogeneous and constant quality of thefinal meat. Additionally, in situ NIRS measurement of intact carcasses

ghts reserved.

at the process line of the slaughterhouse could enable the developmentof simple, fast and objective real-time decision-making support sys-tems. However, a loss of performance has been observed in this kindof NIRS application compared to NIRS analysis of adipose tissues in thelaboratory. While the four major fatty acids (oleic acid, palmitic acid,linoleic acid and stearic acid) could be predicted with a R2 of 96% fromNIR spectra collected from melted subcutaneous adipose tissues in thelaboratory, NIRS analysis through the skin of pigs/carcasses only result-ed in R2 values of 60–77% (in vivo) or 31–94% (carcass) (Pérez-Marín, DePedro, Guerrero-Ginel, & Garrido-Varo, 2009; Pérez-Marín, Garrido-Varo, De Pedro, & Guerrero-Ginel, 2007).

This is partly due to the fact that the light has to travel through theskin (epidermis and dermis) before reaching the target layer (subcuta-neous adipose tissue), which influences the reflectance measurements.The absorption and scatteringproperties of the different tissue layers af-fect the light penetration through the pig skin. Although pioneeringwork has been published about in vivoNIRS applications for pork qualitycontrol (Pérez-Marín et al., 2009), a correct interpretation of measuredreflectance signals still remains absent. Moreover, since the geometry ofthe spectroscopic measurement setup (source and detector properties

Fig. 1. Schematic diagram of a pig skin cross section together with the MC simulated in-strument configuration.

138 E. Zamora-Rojas et al. / Innovative Food Science and Emerging Technologies 22 (2014) 137–146

and source–detector distance) can significantly affect the depth of pen-etration (Arimoto, Egawa, & Yamada, 2005; Inio et al., 2003), this can beoptimized in order tomaximize the sensitivity for photons coming fromthe target layer.

In vitro experiments for determining the light penetration depth arequite difficult since an experiment with sliced tissue samples with suc-cessive thickness would not be very representative. However, with theuse of simulation methods, modeling the transport of light in turbidmedia, this becomes feasible (Arimoto et al., 2005; Inio et al., 2003).Monte Carlo (MC) simulations have been widely used for numericalsimulation of photon propagation in strongly scattering and absorbingmedia, such as biological tissues. MC simulations enable a rigorous,but flexible, numerical solution for the radiative transfer equation(RTE) (Wang& Jacques, 1992). Basically, thismethod launches a photonpacket with an initial weight of unity perpendicularly to the tissue sur-face along the direction of the light beam. Then, a step size is chosen sto-chastically and some weight is lost by the photon packet at the end ofeach step due to the absorption properties of the system. The photonpacket, with the remaining weight, is scattered and a new direction isstochastically determined by sampling from a probability density func-tion, known as the scattering phase function. Finally, when the photonweight is less than a critical threshold, a form of “Russian roulette” isused to determine if the photon packet ends or continues propagating(Tuchin, 2007).

In this study, a Monte Carlo method was consulted to simulate theaverage photon visit depth and fraction of energy absorbed in the NIRregion as a function of the source–detector distance in non-invasivediffuse reflectance/interactance measurements on pig skin. Moreover,experimental spatially resolved reflectance spectroscopy (SRS) mea-surements were performed to show the diffuse reflected light at differ-ent distances from the illumination point and to verify the numericallysimulated results.

2. Materials and methods

2.1. Structure and characteristics of pig skin

Pig skin is composed of three major layers with different structures:epidermis (blood free layer), dermis (vascularized layer with denseirregular connective tissue with collagenous fibers) and hypodermis(adipose tissue composed of two sub-layers separated by thin connec-tive tissue) (Fig. 1). Renaudeau, Leclercq-Smekens, and Herin (2006)reported that Creole pigs, a crossbreed of Iberian and Asian pigs(Lemus-Flores, Ulloa-Arvizu, Ramos-Kuri, Estrada, & Alonso, 2001),had an average epidermis and dermis thickness of respectively 55 μmand 3.82mm. The thickness of the hypodermis or subcutaneous adiposetissue is, however, dependent on the body site. Knowledge of the thick-ness of each layer is important in order to get accurate simulations of thelight penetration. For Iberian pigs, the total subcutaneous adipose tissuelayer or hypodermis is 51 mm thick on average, depending on thebreeding, feeding regime and body site (Daza, Ruiz-Carrascal, Olivares,Menoyo, and López-Bote, 2007; Lyndgaard, Sørensen, van den Berg,and Engelsen, 2011). The outer adipose sub-layer has an average thick-ness of 25 mm in the tail insertion area, while the inner sub-layer was26 mm thick (Dobao et al., 1987).

Knowledge of the bulk optical properties for each individual layer isessential in order to simulate the light propagation through the pigskin. Several authors have optically characterized pig skin, reviewed inBashkatov, Genina, and Tuchin (2011). However, the bulk of the opticalproperties (absorption coefficient (μa), scattering coefficient (μs) and an-isotropy factor (g)) for a wide NIR wavelength range (1150–2250 nm)was only reported by Zamora-Rojas et al. (2013a, 2013b) for each layer(epidermis, dermis and adipose tissue). In Fig. 2 these parameters forthe wavelength range 1150–1850 nm are shown since for longer wave-length values they could not perform all the optical property estimations.Total reflectance, total transmittance and unscattered transmittance

signals were quite low for the case of epidermis and dermis (Zamora-Rojas, Aernouts, Garrido-Varo, Saeys, et al., 2013b). Apart from the bulkoptical properties, the refractive indices of the different layers also haveto be known to simulate the effects of refraction and reflection at theboundaries. Ding, Lu, Jacobs, and Hu (2005) reported average refractiveindex values of 1.41 for pork epidermis and 1.36 for pork dermis forthe studied wavelength range. The refractive index of the subcutaneousadipose tissue was found to be wavelength dependent and can be ap-proximated by the equation (Eq. 1) (Cheng, Shen, Zhang, Huand, andHuang, 2002):

n λð Þ ¼ Aþ Bλ−2 þ Cλ−4 ð1Þ

where λ is the wavelength in nanometers (nm) and A, B and C are differ-ent Cauchy coefficients. For porcine adipose tissue, Cauchy coefficients ofA= 1.4753, B= 4.3902 × 10−3 and C= 0.92385 × 10−9 were derived(Cheng et al., 2002).

2.2. Monte Carlo simulation

A Monte Carlo (MC) method (Arimoto et al., 2005; Inio et al., 2003)was used for the simulation of light propagation through pig skin in the1150–1850 nmwavelength range. The thicknesses, refractive index andbulk optical properties of the different layers are used as input parame-ters for the model. These parameters were described in Section 2.1.

In this study, the simulated optical design consists of a source and adetector arranged concentrically with the source in the core and detec-tor at the annulus (Fig. 1). The simulation was designed as such thatthe detector area captures only the photons whose incidence angle issmaller than 11.5°, corresponding to an optical fiber with numerical ap-erture of 0.2. Source and detector ringswere simulatedwith a diameter/thickness of 600 μm, while the center-to-center distance was variedfrom 0.6 to 5.4 mm in steps of 0.6 mm.

Since the Monte Carlo simulation is a stochastic method, a largenumber of photons are required to obtain reliable results. This makesit computationally intensive and thus a time-consuming procedure. Inthis study, 1,000,000 photons were launched.

139E. Zamora-Rojas et al. / Innovative Food Science and Emerging Technologies 22 (2014) 137–146

2.3. Spatially resolved spectroscopic measurements of pig skin

2.3.1. Spatially resolved spectroscopy setupA translation stage (ThorLabs Inc., New Jersey, USA) setup for spa-

tially resolved spectroscopy (Fig. 3) was used for non-invasive NIRpig skinmeasurementswith different distances between source and de-tector. An Avalight 10 W tungsten halogen light source (Avantes BV.,Apeldoorn, The Netherlands) was coupled to a fiber optical switch(Leoni Fiber Optics GmbH., Neuhaus-Schierschnitz, Germany), whichwas used to switch between sample and reference measurements. A

1200 1300 14000

5

10

15

20

25

30a

b

Wavel

1200 1300 140090

100

110

120

130

140

150

160

170

180

190

200

Wavel

µ s (

cm−1

)µ a

(cm

−1)

Fig. 2. Average optical properties for each pig skin layer reported by Zamora-Rojas et al. (2013acoefficient; c) Anisotropy factor.

neutral density filter (NDF, Qioptiq, Luxembourg) of 3.0 optical densitywas used to prevent detector saturation during the reference mea-surements. A fixed contact probe with a central single multimode fiber(diameter: 600 μm and numerical aperture: 0.22. Romack Inc.,Williamsburg, USA) was used to guide the light from the light source tothe tissue,while a similar parallel probe/fiber collects the diffuse reflectedlight. The detector fiber can be moved at various precise distances fromthe illuminating fiber, with minimal and maximal distances betweenthe fiber centers of respectively 1.1 and 25 mm. The collected light isthen guided to a diode array spectrometer working in the 400–1650 nm

1500 1600 1700 1800

ength (nm)

EpidermisDermisSubcutaneous adipose tissue

1500 1600 1700 1800

ength (nm)

EpidermisDermisSubcutaneous adipose tissue

, 2013b) in the wavelength range 1150–1850 nm. a) Absorption coefficient; b) Scattering

1200 1300 1400 1500 1600 1700 18000.88

0.9

0.92

0.94

0.96

0.98

1

Wavelength (nm)

g

EpidermisDermisSubcutaneous adipose tissue

c

Fig. 2 (continued).

140 E. Zamora-Rojas et al. / Innovative Food Science and Emerging Technologies 22 (2014) 137–146

spectral range (Zeiss Corona VIS NIR 1.7, Carl Zeiss AG., Jena, Germany).Automatic control of the system was obtained through a custom-madeprogram in LabView 8.5 (National Instruments, Austin, TX).

2.3.2. Sample setTwenty-five Iberian pig skin (composed of epidermis, dermis and

subcutaneous adipose tissue) samples with a size of 5 × 5 × 3 cm3

were collected at a Spanish slaughterhouse, frozen in liquid nitrogenand stored at −20 °C until 24 h before the analysis. The optical fibers(source and detector) were located on the center of the sample,touching the pig skin. The spatially resolved reflectance spectra were

Fig. 3. Schematic diagram of the translation stage SRS setup. FS: fiber switch; F: fiber opti

collected for source–detector distances from 1.2 mm to 4.8 mm, witha step of 0.6 mm.

A comparison of these measurements with the simulated resultswas performed for the common wavelength range (1150–1650 nm)and source–detector distances available (1.2–4.8 mm, with a step of0.6 mm). Since the optical configuration of the simulations and theSRS detector differs, a scaling factor was applied to the SRS profiles tobe able to compare the results. This scaling factor was calculated asthe ratio between the ring-shaped detector area of the MC simulationfor each source–detector distance and the area of the SRS point-detection fiber (0.00282 cm2).

c probe; NDF: neutral density filter; TS: translation stage; DAQ: data acquisition card.

1200 1300 1400 1500 1600 1700 18000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Wavelengths (nm)

Fra

ctio

n of

abs

orbe

d en

ergy

at

each

laye

r

EpidermisDermis

Hypodermis (1st layer)

Hypodermis (2nd layer)

Fig. 5. Fraction of absorbed energy at each layer for thewavelength range 1150–1850 nm.Hypodermis values have been multiplied by a factor of 10 times for the graph.

141E. Zamora-Rojas et al. / Innovative Food Science and Emerging Technologies 22 (2014) 137–146

3. Results and discussion

3.1. Monte Carlo simulation of light propagation in pig skin

In Fig. 4 the average photon visit depth within the detected photonsis shown. This was calculated as the average rate of energy at a gridelement in the simulation which is divided in a new direction for itspropagation (Inio et al., 2003). A decrease in the average photon visitdepth was observed at the wavelength range around 1450 nm, whichis related with the strong absorption coefficients in the epidermis anddermis due towater. This indicates the impact of the tissue's absorptioncoefficients on the light propagation.Moreover, for source–detector dis-tances of 3.6mmormore,MC simulations failed to estimate the averagephoton visit depth at the above mentioned water-related absorptionband as the absorption was too high and consequently no photonsreached the surface. The penetration depth, or average photon visitdepth, is maximum in the range around 1150 and 1270 nm, wherethe absorption by water and lipids is minimal for the consideredwavelength range. In general, the average photon visit depth increasedwith increasing source–detector distance. However, even for thelargest source–detector distance, the average photon visit depth stillcorresponded to a position in the dermis layer.

In Fig. 5 the fraction of injected photons (sourcefiber) that is absorbedby each layer is illustrated. Themajority of absorption (70–90%) occurredin the dermis, although the absorption of photons bywater (1440 nm) inthe epidermis resulted in a significant reduction. Only a small number ofphotons in the 1150–1340 nmand 1550–1850 nm range reached the ad-ipose tissue layer and could be absorbed there.

In Fig. 6 the percentage of energy captured by the detector at eachlayer as a function of the source–detector distance is shown, it meansthe percentage of the photons coming from each layer captured bythe detector compared to the total energy captured. The epidermis(Fig. 6a) contributed with almost 3% of the energy captured in the1150–1400 nm range to the signal, independent of the source–detectordistance. At longer wavelengths, the fraction of captured energy washigher and increased as the source–detector distance increased. At1580–1850 nm, a slight increase from 7 to 11% on average was noticed

1200 1300 1400 10

0.5

1

1.5

2

2.5

Wavele

Ave

rage

pho

ton

visi

t de

pth

(mm

)

Fig. 4. Average photon visit depth for d

from the shortest compared to the longest source–detector distance.The dermis layer (Fig. 6b) sent most of the energy (+/−94%), exceptfor the water-related band around 1440 nmwhere most of the absorp-tion took place in the epidermis. However, as the source–detector dis-tance increased, a slight decrease in the percentage of captured energywas observed. In Fig. 6c the simulation results are shown for the cap-tured energy as a function of the source–detector distance for theouter adipose tissue layer. As the source–detector distances increase,the contribution of the first subcutaneous adipose layer to the capturedfraction increased from 0.3% to 3% at around 1150–1340 nm. No pho-tons reached the detector coming from the second sub-layer (Fig. 6d).

For larger source–detector distances, the average photon visit depthwas larger and the contribution of the subcutaneous adipose tissue(i.e. target layer containing fatty acid profile information) to the detected

500 1600 1700 1800

ngth (nm)

0.6 mm1.2 mm1.8 mm2.4 mm3.0 mm3.6 mm4.2 mm4.8 mm5.4 mm

ifferent source–detector distances.

142 E. Zamora-Rojas et al. / Innovative Food Science and Emerging Technologies 22 (2014) 137–146

signal was higher. Nevertheless, the signal was dominated in all thecases by the structure and composition of the dermis layer. As NIRmea-surements collected in vivo for fatty acid prediction in Iberian pigsshowed suitable performance for the on-farm application (Pérez-Marín et al., 2009), the fatty acid information in the acquired spectramight have come from the fat in the dermis instead of the subcutaneousadipose tissue. Nevertheless, the dermis layer contains fat which com-position may be related somehow with the one present in the fat ofthe subcutaneous adipose tissue since the quantitative models showeda suitable performance.

Another important observation was that by increasing the source–detector distance, a larger fraction of the collected photons had traveleddeeper. However, the diffuse reflectance signal became low very fast. InFig. 7 the simulated diffuse reflectance spectra captured by the annulusdetector as a function of the source–detector distance are illustrated. As

115012400.6

1.21.8

2.43.0

3.64.2

4.85.4

0

10

20

30

40

50

Source−detector distance (mm)

Cap

ture

d en

ergy

(%

)

Epide

115012400.6

1.21.8

2.43.0

3.64.2

4.85.475

80

85

90

95

100

Source−detector distance (mm)

Cap

ture

d en

ergy

(%

)

De

a

b

Fig. 6. Percentage of captured energy as a function of the source–detector distance. a) Epiderm(2nd layer).

it can be seen, the reflectance spectra in the 1150–1850 nm range weredominated by water absorption. As the source–detector distance in-creased, a steady decline in the diffuse reflectance values could be no-ticed. For wavelengths above 1350 nm, where the absorption is high,the reflectance spectra are nearly 0 (b0.1%) for distances larger than2.4 mm. These results are relevant for NIR sensor developers interestedin non-invasive applications as these results showed how the capturedspectra change in relation to the properties of the measured product.

3.2. Spatially resolved diffuse reflectance NIR spectra of pig skin

In order to validate the MC simulations, spatially resolved reflec-tance spectra were collected on 25 Iberian pig skin samples. In Fig. 8athe average spectra in the visible and NIR regions (450–1650 nm)mea-suredwith the SRS system are shown,while Fig. 8b shows a zoom to the

13401440

15401650

17501850

Wavelength (nm)

rmis

13401440

15401650

17501850

Wavelength (nm)

rmis

is; b) Dermis; c) Subcutaneous adipose tissue (1st layer); d) Subcutaneous adipose tissue

11501240

13401440

15401650

17501850

0.61.2

1.82.4

3.03.6

4.24.8

5.40

50

100

Cap

ture

d en

ergy

(%

)

05.4

4.84.2

3.63.0

2.41.8

1.20.6 1150

12401340

1440

Wavelength (nm)1540

1650 17501850

3

2

1

4c

d

Cap

ture

d en

ergy

(%

)

Wavelength (nm)

Source−detector distance (mm)

Source−detector distance (mm)

Subcutaneous adipose tissue (2nd layer)

Fig. 6 (continued).

143E. Zamora-Rojas et al. / Innovative Food Science and Emerging Technologies 22 (2014) 137–146

wavelength range common with the MC simulations (1150–1650 nm)after scaling the reflectance values, as described in Section 2.3.

The SRS device measures the reflectance/interactance with a singleoptical fiberwith an area of 0.00282 cm2, while theMC simulations con-sider an annular detector configuration surrounding the point source(Fig. 1). Therefore, the MC detector area increased with the increasingdistance between the source and the detector.

Similar to theMC simulation results, a decline in the SRS reflectancevalues was observed when the source–detector distance increased. Re-flectance spectrameasured on Iberian pig skin sampleswere dominatedby absorption bands related to deoxy-hemoglobin and oxy-hemoglobinin the visible region (560 and 580 nm, respectively) and water in theNIR range (around 900, 1200 and 1400–1580 nm) (Fig. 8a).

In Figs. 7 and 8b the MC simulated and scaled experimental SRS dif-fuse reflectance spectra in the NIR range are illustrated for the in vitro

Iberian pig skin samples, respectively. A visual analysis of the spectralshape in the common wavelength range (1150–1650 nm) shows agood agreement between both figures. However, the absorption bandsrelated towaterwere broader for the case of the experimentalmeasure-ments. A possible explanation for this might be the higher spectral res-olution (6.2 nm) used for the measurements of the bulk opticalproperties of the tissues (Aernouts et al., 2013) which have been usedin the simulations, compared to the spectral resolution (18 nm) of thespectrometer in the SRS setup. In Fig. 9 theMC simulated andmeasuredspectra at a source–detector distance of 2.4 mm are illustrated fromwhich the above comments can be clearly observed. A good agreementwas found between the simulated and measured spectra for wave-lengths longer than 1300nm. Below1300 nm, themeasured reflectancevalues are lower than the simulated values. A possible explanation forthis mismatch might be a difference in the bulk optical properties

11501240

13401440

15401650

17501850

5.44.8

4.23.6

3.02.4

1.81.2

0.60

0.05

0.1

0.15

0.2

Wavelength (nm)

Source−detector distance (mm)

Ref

lect

ance

(%

)

Fig. 7.Monte Carlo estimated diffuse reflectance spectra of an Iberian pig skin sample with average bulk optical properties (Fig. 2) as a function of the source–detector distance.

600800

10001200

14001600

1.21.8

2.43

3.64.2

4.8

0

0.01

0.02a

b

Wavelength (nm)

Source−detector distance (mm)

Source−detector distance (mm)

Ref

lect

ance

(%

)

Ref

lect

ance

(%

)

0.1

1.21.8

2.43.0

3.54.2

4.8 11501250

1350

Wavelength (nm)1450

15501650

0.08

0.06

0.04

0.02

0

Fig. 8. Average spatially resolved diffuse reflectance spectra of Iberian pig skin for different source–detector distances. a) Visible and NIR regions; b) NIR region (scaled values to considerthe same optical configuration as in the MC simulations).

144 E. Zamora-Rojas et al. / Innovative Food Science and Emerging Technologies 22 (2014) 137–146

1150 1200 1250 1300 1350 1400 1450 1500 1550 1600 16500

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1x 10−3

Wavelength (nm)

Ref

lect

ance

(%

)

SRS profileMC profile

Fig. 9. Comparison of the average MC simulated and SRS measured spectra of Iberian pig skin in the wavelength range 1150–1650 nm for the 2.4 mm source–detector distance.

145E. Zamora-Rojas et al. / Innovative Food Science and Emerging Technologies 22 (2014) 137–146

and/or thickness of the different tissue layers for themeasured samples,compared to the average values from literature used in the simulations.Another possible explanation might be an imperfect contact betweenthe sample and the source and detector fibers. In the simulations, a per-fect contact was assumed while it is likely that there may have been avery thin layer of air in between the tissue and the fibers resulting in arefractive indexmismatch and extra reflections. The effect of the criticaldetection angle on the diffuse reflectance values should, however, belimited since both SRS and MC simulated detectors had practically thesame numerical aperture. Nevertheless, other parameters such as aver-age photon visit depth or the fraction of absorbed energy at each layer,discussed in the above section, are independent of the numerical aper-ture of the detector since light reaching the detector is scattered almostisotropically after many scattering events (Inio et al., 2003). These mea-surement results validate the in vitro pig skin observations presentedand discussed in this study based on a Monte Carlo method for simula-tion of light propagation in multi-layered biological tissues.

4. Conclusions

In this study, the effect of the source–detector distance on the NIRdiffuse reflectance spectra acquired from Iberian pig skin samples wasinvestigated. Monte Carlo simulations of light propagation showedthat large source–detector distances enabled going deeper in the skintissue, maximizing the sensitivity of the captured signal to the subcuta-neous adipose tissue. However, the results for the average photon visitdepth and fraction of absorbed energy indicated that the dermis layerwas responsible for most of the absorption. A compromised decisionshould be taken in order to select the optimal source–detector distancessince large values enabled going deeper in the tissue, but the simulatedand experimental SRS diffuse reflectance spectra showed a serious re-duction in the signal and with that the signal-to-noise ratio. Further re-search focusing on the 1100–1450 nm wavelength range might beinteresting as water absorption is not as dominant there, revealing aspectral window for the detection of other chemical or physical infor-mation. Moreover, other instrumental design could be investigated toget better SNR with sufficient information from the deeper layers.

Acknowledgments

Eduardo Zamora-Rojas acknowledges the financial support of theSpanish Ministry of Education as a fellow of the Program “Training ofUniversity Teachers” (Formación del Profesorado Universitario, FPU).Ben Aernouts is funded as aspirant fellow of the Research Foundation—Flanders (Brussels, Belgium). This work was financed by the researchprojects: SB-090053 (GlucoSens project, I.W.T.—Flanders) and P09-AGR-5129 (Research Excellence Program of the Andalusian RegionalGovernment). The authors are grateful to the Spanish ProtectedDesigna-tion of Origin Los Pedroches and to the Spanish Iberian pork industryCOVAP S.C.A. for providing the experimental materials.

References

Aernouts, B., Zamora-Rojas, E., Van Beers, R., Watté, R., Wang, L., Tsuta, M., et al. , (2013).Supercontinuum laser based optical characterization of intralipid phantoms in the500–2250 nm range. Optics Express, 21, 32450–32467.

Arimoto, H., Egawa, M., & Yamada, Y., (2005). Depth profile of diffuse reflectancenear-infrared spectroscopy for measurement of water content in skin. Skin Researchand Technology, 11, 27–35.

Bashkatov, A. N., Genina, E. A., & Tuchin, V. V., (2011). Optical properties of skin, subcuta-neous and muscle tissues: a review. Journal of Innovative Optical Health Sciences, 4(1),9–38.

Batten, G., (2012). Special issue: medical application. Journal of Near Infrared Spectroscopy,20(1).

Cheng, S., Shen, H. Y., Zhang, G., Huand, C. H., & Huang, X. J., (2002). Measurement of therefractive index of biotissue at four laser wavelengths. Proceedings of SPIE, 4916,172–176.

Daza, A., Ruiz-Carrascal, J., Olivares, A., Menoyo, D., & López-Bote, G. J., (2007). Fatty acidsprofile of the subcutaneous backfat layers from Iberian pigs raised under free-rangeconditions. Food Science and Technology International, 13(2), 126–140.

Ding, H., Lu, J. Q., Jacobs, K. M., & Hu, X., (2005). Determination of refractive indices of por-cine skin tissues and intralipid at eight wavelengths between 325 and 1557 nm.Journal of the Optical Society of America, 22(6), 1151–1157.

Dobao, M. T., Rodrigañez, J., Silio, L., Toro, M.A., De Pedro, E., & García de Siles, J. L., (1987).Crecimiento y características de canal en cerdos ibéricos, duroc-jersey × ibérico yjiaxin × ibérico. Investigación Agraria, 2(1), 9–23.

Inio, K., Maruo, K., Arimoto, H., Hyodo, K., Nakatani, T., & Yamada, Y., (2003). Monte Carlosimulation of Near Infrared Reflectance Spectroscopy in the wavelength range from1000 nm to 1900 nm. Optical Review, 10(6), 600–606.

Lemus-Flores, C., Ulloa-Arvizu, R., Ramos-Kuri, M., Estrada, F. J., & Alonso, R. A., (2001).Genetic analysis of Mexican hairless pig populations. Journal of Animal Science, 76,3021–3026.

146 E. Zamora-Rojas et al. / Innovative Food Science and Emerging Technologies 22 (2014) 137–146

Lyndgaard, L. B., Sørensen, K. M., van den Berg, F. V., & Engelsen, S. B., (2011). Depth pro-filing of porcine adipose tissue by Raman spectroscopy. Journal of RamanSpectroscopy, 43(4), 482–489.

Pérez-Marín, D., De Pedro, E., Guerrero-Ginel, J. E., & Garrido-Varo, A., (2009). Afeasibility study on the use of near-infrared spectroscopy for prediction ofthe fatty acid profile in live Iberian pigs and carcasses. Meat Science, 83,627–633.

Pérez-Marín, D., Garrido-Varo, A., De Pedro, E., & Guerrero-Ginel, J. E., (2007).Chemometric utilities to achieve robustness in liquid NIRS calibrations: ap-plication to pig fat analysis. Chemometrics and Intelligent Laboratory Systems,87, 241–246.

Renaudeau, D., Leclercq-Smekens, M., & Herin, M., (2006). Differences in skin characteris-tics in European (Largewhite) and Caribbean (Creole) growing pigswith reference tothermoregulation. Animal Research, 55, 299-217.

Tuchin, V. V., (2007). Methods and algorithms for themeasurement of optical parametersof tissues. In V. V. Tuchin (Ed.), Tissue Optics: Light Scattering Methods and Instrumentsfor Medical Diagnosis (pp. 143–256) (2nd ed.). Washington, USA: SPIE.

Wang, L., & Jacques, S. L., (1992). Monte Carlo modeling of light transport inmulti-layeredtissues in standard C. Available at: http://labs.seas.wustl.edu/bme/wang/mcr5/mcman.pdf (Accessed: 12 November 2012)

Zamora-Rojas, E., Aernouts, B., Garrido-Varo, A., Guerrero-Ginel, J. E., Pérez-Marín, D., &Saeys, W., (2013a). Double integrating spheres measurements for estimating opticalproperties of pig subcutaneous adipose tissue. Innovative Food Science and EmergingTechnologies, 19, 218–226.

Zamora-Rojas, E., Aernouts, B., Garrido-Varo, A., Saeys, W., Pérez-Marín, D., &Guerrero-Ginel, J. E., (2013b). Optical properties of pig skin epidermis and dermis es-timated with double integrating spheres measurements. Innovative Food Science andEmerging Technologies, 20, 343–349.