www. biophotonics-journal.orgJournal of

BIOPHOTONICS

REPRINT

LETTER

2-D mapping of skin chromophoresin the spectral range 500–700 nm

Dainis Jakovels* and Janis Spigulis

Bio-Optics and Fibre Optics Laboratory, Institute of Atomic Physics and Spectroscopy, University of Latvia, Raina Blvd. 19,LV – 1586, Riga, Latvia

Received 17 August 2009, revised 20 October 2009, accepted 20 October 2009Published online 9 November 2009

Key words: multi-spectral imaging, haemoglobin, melanin, skin chromophore mapping

PACS: 00.00.Xx, 11.11.Yy

# 2010 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

1. Introduction

The mapping of in-vivo skin chromophores is basedon multi-spectral imaging that combines spectralanalysis of diffusely reflected light and image analy-sis, resulting in 2-D maps of the relative concentra-tions of chromophores, e.g. oxy-/deoxy-haemoglobinand melanin [1, 2]. Such mapping ensures reliable

non-invasive evaluation of skin condition [3–5]. Theleast-squares regression analysis of a broad visible-NIR spectral range (450–720 nm [2], 460–820 nm[6]) was successfully used to estimate the chromo-phore content in skin. Drawback of this technique isvery time-consuming (proportional to the quantityof spectral information) data acquisition and proces-sing; narrowing of the working spectral range cer-

# 2010 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Journal of

BIOPHOTONICS

The multi-spectral imaging technique has been used fordistant mapping of in-vivo skin chromophores by analyz-ing spectral data at each reflected image pixel and con-structing 2-D maps of the relative concentrations ofoxy-/deoxy-haemoglobin and melanin. Instead of using abroad visible-NIR spectral range, this study focuses onnarrowed spectral band 500–700 nm, speeding-up thesignal processing procedure. Regression analysis con-firmed that superposition of three Gaussians is optimalanalytic approximation for the oxy-haemoglobin absorp-tion tabular spectrum in this spectral band, while super-position of two Gaussians fits well for deoxy-haemoglo-bin absorption and exponential function – for melaninabsorption. The proposed approach was clinically testedfor three types of in-vivo skin provocations: ultravioletirradiance, chemical reaction with vinegar essence andfinger arterial occlusion. Spectral range 500–700 nm pro-

vided better sensitivity to oxy-haemoglobin changes andhigher response stability to melanin than two reducedranges 500–600 nm and 530–620 nm.

Relative oxy-hemoglobin concentration maps.

* Corresponding author: e-mail: dainis.jakovels@gmail.com, Phone: +37 167 225 493, Fax: +37 167 228 249

J. Biophoton. 3, No. 3, 125–129 (2010) / DOI 10.1002/jbio.200910069

tainly could speed up the procedure. However, thereis a risk to loose specificity and sensitivity to themain skin chromophores, therefore optimal spectralrange for skin chromophore mapping should befound [7]. For instance, potential of the reducedspectral range 525–645 nm for imaging of skin hae-moglobin oxygen saturation has been demonstratedrecently [8]. The third main skin chromophore –melanin – could not be mapped since the measure-ments were taken for palm skin.

Goal of the present study was to examine spec-tral interval 500–700 nm and two sub-bands 500–600 nm and 530–620 nm from the point of applic-ability for simultaneous distant mapping of threemain in-vivo skin chromophores: oxy-haemoglobin,deoxy-haemoglobin and melanin. Feasibility of thisapproach was checked by pilot multi-spectral meas-urements of provoked skin.

2. Experimental

The multi-spectral imaging system Nuance 2.4 (Cam-bridge Research & Instrumentation, Inc., USA) anda PC were used for spectral imaging of normal andprovoked in-vivo skin areas of the forearm. Illumi-nation source was a 100 W tungsten incandescentlamp (intensity fluctuations less than �2% duringthe measurement time) with linear polarization filter.This polarizer was oriented orthogonally to thebuilt-in polarizer of Nuance 2.4, so significantly redu-cing the influence of skin specular reflection [9].

The system was adjusted for spatial resolution0.75 � 0.75 mm (the pixel size) and spectral resolu-tion 10 nm (bandwidth of the Nuance 2.4 liquid crys-tal tuneable filter).

2.1 Data acquisition and processing

The data were collected in an image cube – a stackof intensity images at numerous wavelength bands.Typical time required for creation of the image cubein spectral interval 500–700 nm was �10 s.

The back reflected light intensity (I) values ateach pixel were transformed to the optical density(OD) as follows:

OD ¼ �log10II0

� �; ð1Þ

where I0 – reflection intensity from the white refer-ence – bended white office paper sheet (spectral re-flectance 0.90 � 0.04 within the 500–700 nm band),attached to the forearm skin. Optical density of the

superficial skin layer has been predicted in frame ofthe three chromophore absorption model:

ODpredicted ¼ aOH � eOH þ aDOH � eDOH

þ aMel � eMel þ aOffset ; ð2Þwhere eOH, eDOH and eMel denote reference absorp-tion spectra for oxy-hemoglobin (OH), deoxy-hemo-globin (DOH) and eumelanin as melanin (Mel), re-spectively. aOH, aDOH and aMel represent the relativechromophore concentration values; aOffset is the dif-ference between the predicted and measured spec-tra.

The predicted OD spectrum at each image pixelwas compared to the measured OD spectrum by sol-ving the nonlinear least-squares problem using theTrust-Region algorithm [10], with subsequent extrac-tion of the corresponding relative concentrations ofthe skin chromophores [2]. The reference absorptionspectra of the three chromophores were taken fromthe literature data [11, 12].

Analysis of the reference spectra allowed propos-ing handy analytic expressions that approximatedwell the tabular data within the spectral interval500–700 nm. Superposition of three Gaussiansproved to be optimal for approximation of the OHspectrum, while superposition of two Gaussians suit-ed well for approximation of the DOH spectrum(Figure 1). The values of multiple determination

Figure 1 (online color at: www.biophotonics-journal.org)Analytical approximations of the tabular spectra of hae-moglobin [12] in the 500–700 nm range: (a) superpositionof three Gaussians related to the OH spectrum, (b) super-position of two Gaussians related to the DOH spectrum.

D. Jakovels and J. Spigulis: 2-D mapping of skin chromophores in the spectral range 500–700 nm126

Journal of

BIOPHOTONICS

# 2010 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.biophotonics-journal.org

coefficients R2 were obtained as R2OH ¼ 0.9988 for

oxy-haemoglobin and R2DOH = 0.9973 for deoxy-hae-

moglobin. Generally, the Gaussian superposition canbe expressed as:

f ðxÞ ¼ a1 � e�

x�b1c1

�þ a2 � e

x�b2c2

� �þ an � e

�x�bn

cn

�; ð3Þ

where a, b, c – analytically expressed coefficientsand x – wavelength.

Regarding the melanin absorption spectrum, itcould be well approximated (R2

Mel ¼ 0.9975) by theexponential function:

f ðxÞ ¼ a � e�b � x ; ð4Þwhere a, b – positive coefficients.

As the next step, the relative values of the re-spective skin chromophore concentrations aOH,aDOH, aMel have been determined in MatLab at eachimage pixel, and the chromophore maps represent-ing the planar distribution of the particular chromo-phore have been constructed. The color scale (Fig-

ures 2, 3) represent the obtained chromophoreconcentrations relatively to their mean values of thesurrounding normal (unprovoked) skin.

2.2 Skin provocations

Three different provocations of skin photo-type 3(single volunteer) were applied in order to obtain pi-lot results for the proposed method.

Electrodeless high-frequency discharge Mercurylamp with specific UV-C peak at 253.7 nm was usedto irradiate the forearm skin through a mask with1� 1 cm apertures for local irritation. Four differentdoses (1, 2, 3 and 4 minutes provocation time) wereapplied to achieve different skin responses. Erythe-ma appeared at all provocation areas in-between0.5 hours, and skin colour changed from dark red todark brown in the next days. Five measurement ser-ies were taken during the first day – 10, 30, 60, 120and 240 minutes after the provocation, and the mea-surements were repeated 1, 2, 3, 4, 7 and 10 daysafter the provocation.

Vinegar essence was used for chemical skin pro-vocation. Two stripes (7 mm width) were applied for3 and 5 minutes to achieve different forearm skin re-action intensities. Skin erythema appeared in fewminutes. Measurement series were taken 10, 30 60and 90 minutes after the provocation.

Finally, a resin cuff for arterial occlusion (p �150 mm Hg) was used to reduce skin blood oxygena-tion in a finger [13]. Measurements were taken dur-ing full occlusion and immediately after removal ofthe cuff.

After appropriate signal processing, the maps ofrelative chromophore concentrations were createdto follow-up the provoked skin responses.

3. Results and discussion

3.1 Comparison of spectral ranges

The spectral range 500–700 nm clearly showed bettersensitivity to the OH content changes and higher sta-bility to melanin if compared to the narrower bands500–600 nm and 530–620 nm. Sensitivity to the OHcontent changes was evaluated as contrast betweenprovoked area and normal skin, and for the range500–700 nm it was for �20% higher than that forboth reduced spectral ranges. Stability to melaninwas verified by analysing the chemical and mechani-cal provocations where its concentration increase wasnot expected; haemoglobin changes in these testscould influence results causing melanin artefacts.

Figure 2 (online color at: www.biophotonics-journal.org)Parameter maps for 4 minutes UV-C provocation 30 min-utes (left) and 4 days (right) after irritation. The arrow atthe DOH and melanin maps is a pen marker.

J. Biophoton. 3, No. 3 (2010) 127

LETTERLETTER

# 2010 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheimwww.biophotonics-journal.org

False-increased melanin content at the chemical pro-vocation areas was obtained using the narrower spec-tral ranges 500–600 nm (up to 20%) and 530–620 nm(up to 30%). Consequently, the spectral interval 500–700 nm was chosen as the best option for simulta-neous mapping of the three skin chromophores.

3.2 UV-provocation responses

Visible skin erythema appeared within 30 minutes to2 hours at all UV-provoked areas where increasedconcentrations of OH (up to 300% compared to nor-mal skin) and DOH (up to 50%) were obtained. Re-sponse depended on the provocation doses – higherparameter changes at longer irradiation times wereobserved. The obtained OH concentration growthwas faster than that of the DOH concentration; itreached maximum values (from þ300% for 1 min.dose to þ400% for 4 min.) within the first day afterprovocation. High increase of the OH and DOH con-centrations (up to þ400%) remained for the next 4days. Later the concentrations gradually returned tothe normal level. Slower increase of the parametervalues was observed for DOH, but it remained high(from þ20% for 1 min. dose to þ100% for 4 min.)for a week. Increased melanin concentration ap-peared on the second day (þ5%) after UV-provoca-tion. Higher parameter increase was observed forhigher provocation doses (from þ5% for 1 min. doseto þ20% for 4 min.). In its terms, the results corre-sponded with previous UV-provocation studies [14].

For illustration, skin chromophore maps relatedto the UV-provocation (4 minutes) are presented atFigure 2.

3.3 Chemical provocation responses

Visible skin erythema appeared within several min-utes after the chemical provocation. Increased OHconcentration at the irritated area has been ob-tained, with maximum �30 minutes after provoca-tion (+200% for 3 min. dose, þ300% for 5 min.dose); more pronounced response corresponded tolonger irritation time. Slight increase of DOH con-centration (less than 15%) at the chemically pro-voked area was obtained, as well.

3.4 Responses during and after the fingerocclusion

Before occlusion the middle finger maps looked likethese for the neighbour fingers. Decreased OH con-

centration (down to 10%) 5 minutes after the arte-rial finger occlusion was obtained (Figure 3). Rightafter the cuff release it increased up to 400% rela-tively to the level before occlusion. Thus the ex-pected OH decrease and overshoot was observed[13]. Slight decrease of DOH (�20%) was obtainedduring the occlusion. The background was a woodentable with reflectance spectrum similar to the skinspectrum (correlation coefficient 0,86), thereforealso the background part of images was processed.However, the intensity distribution of backgroundwas about the same at both images (signal standarddeviation over all pixels did not differ), which con-firms adequate processing.

4. Conclusions

In frame of the simplified 3-chromophore model, thespectral range 500–700 nm is considered to be opti-mal for simultaneous mapping of OH, DOH andmelanin; narrowing of this range (500–600 nm, 530–620 nm) has lead to unacceptable results.

Tabulated molar absorption spectral data for thisregion can be well approximated by superposition ofthree Gaussians for OH, superposition of two Gaus-sians for DOH and exponential function for melanin.Such analytical approximations considerably reducedthe signal processing time.

Efficiency of the proposed model and methodol-ogy was confirmed by the test measurements. Threedifferent provocations (UV, chemical and arterial oc-clusion) resulted in notable changes of the skin chro-mophore content and were well reflected in the ob-tained maps of the three main chromophores.

The accuracy of chromophore mapping can befurther improved taking into account more specific as-pects like different light penetration depths at variouswavelengths [15] and the scattering effects in skin [16].Experimental comparison with other skin mappingmethods [17] would be performed. There is still poten-

Figure 3 (online color at: www.biophotonics-journal.org)Relative OH concentration maps after the finger occlusion(Cuff on) and immediately after release (Cuff off).

D. Jakovels and J. Spigulis: 2-D mapping of skin chromophores in the spectral range 500–700 nm128

Journal of

BIOPHOTONICS

# 2010 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.biophotonics-journal.org

tial to speed-up the mapping procedure by modifyingthe data processing algorithms. Only pilot results havebeen reported here; essentially more experimentaldata are needed for approbation of the new method.

References

[1] G. N. Stamatas and N. Kollias, In Vivo documentationof cutaneous inflammation using spectral imaging. J.Biomed. Opt. 12(5), 051603 (2007).

[2] M. A. Ilias, E. Haggblad, and C. Anderson, Visible,hyperspectral imaging evaluating the cutaneous re-sponse to ultraviolet radioation. Proc. SPIE, Vol.6441, 644103 (2007).

[3] L. L. Randeberg, A. M. Winnem, and N. E. Langlois,Skin (Los. Angeles) changes following minor trauma.Laser Surg. Med. 39(5), 403–413 (2007).

[4] C. Balas, G. Themelis, and A. Papadakis, A Novel-Spectral Imaging System: Application on in-vivo De-tection and Grading of Cervical Precancers and ofPigmented Skin Lesions. In Proc. of “Computer Vi-sion Beyond the Visible Spectrum” CVBVS’01 Work-shop, Hawaii, USA, Dec. (2001).

[5] A. Vogel, V. V. Chernomordik, and S. G. Demos,Using noninvasive multispectral imaging to quantita-tively assess tissue vasculature. J. Biomed. Opt. 12(5),051604 (2007)

[6] G. Zonios, J. Bykowski, and N. Kollias, Skin, melanin,haemoglobin, and light scattering properties can bequantitatively assessed in vivo using diffuse reflec-tance spectroscopy. J. Invest. Dermatol. 117(6), 1452–1457 (2001).

[7] M. E. Eames, J. Wang, and B. W. Pogue, Wavelengthband optimization in spectral near-infrared optical to-mography improves accuracy while reducing data ac-

quisition and computational burden. J. Biomed. Opt.13(5), 054037 (2008).

[8] K. J. Zuzak, M. T. Gladwin, and R. O. Cannon III,Imaging haemoglobin oxygen saturation in sicklecell disease patients using noninvasive visible reflec-tance hyperspectral techniques: effects of nitricoxide. Am J. Physiol. Heart. Circ. Physiol. 285,H1183–H1189, 2003.

[9] S. G. Demos and R. R. Alfano, Optical polarizationimaging. App. Opt. 36(1), 150–155 (1997).

[10] R. H. Byrd, R. B. Schnabel, and G. A. Schultz, Atrust region algorithm for nonlinearly constrained op-timization. SIAM J. Numer. Anal., 24, 1152–1170(1987).

[11] S. Prahl, Tabulated Molar Extinction Coefficientfor Hemoglobin in Water, http://omlc.ogi.edu/spec-tra/hemoglobin/summary.html, Last access: August,2009.

[12] S. L. Jacques, Extinction coefficient of melanin, http://omlc.ogi.edu/spectra/melanin/eumelanin.html, Lastaccess: August, 2009.

[13] D. J. Clark, T. J. H. Essex, and B. Cater, Skin (Los.Angeles) oxygen saturation imager. Adv. Exp. Med.Biol. 428, 573–577 (1997).

[14] P. M. Farr, J. E. Besag, and B. L. Diffey, The timecourse of UVB and UVC erythema. J. Invest. Derma-tol. 91(5), 454–457 (1988).

[15] S. R. Arridge, M. Hiroaka, and M. Schweiger, Statisti-cal basis for the determination of optical pathlengthin tissue. Phys. Med. Biol. 40(9), 1539–1558 (1995).

[16] I. A. Kiseleva and Y. P. Sinichkin, The apparentoptical density of the scattering medium: influenceof scattering, Proc. SPIE, Vol. 4707, 223–227(2002).

[17] L. E. Dolotov, I. A. Kiseleva, and Y. P. Sinichkin, Di-gital imaging of human skin. Proc. SPIE, Vol. 5067,139–147 (2003).

J. Biophoton. 3, No. 3 (2010) 129

LETTERLETTER

# 2010 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheimwww.biophotonics-journal.org