Imaging of hemoglobin oxygen saturation ratio in the

face by spectral camera and its application to evaluate

dark circles

Kumiko Kikuchi, Yuji Masuda and Tetsuji HiraoShiseido Research Center, Yokohama, Japan

Background: Contact-type spectrophotometers have been

widely used to measure skin color to determine the color val-

ues and melanin and hemoglobin contents. Recently, a spec-

tral camera was introduced to evaluate two-dimensional color

distribution. However, its application to skin color measurement

has been limited.

Methods: The original spectral imaging system developed for

facial skin consisted of a spectral camera and an original light-

ing unit for uniform irradiation of the face. The distribution of

skin chromophores in the face, including melanin and oxy- and

deoxyhemoglobin, was calculated from the reflectance data for

each pixel of the spectral images. In addition, to create a mean

spectral image of the group, a face morphing technology for

spectral data was proposed. Using the system, we determined

the characteristics of the dark circles around the eyes and also

evaluated the effects of an anti-dark circle cosmetic.

Results: This system enabled the sensitive detection of skin

chromophores in the face. Melanin content increased and

hemoglobin oxygen saturation ratio decreased locally in the

infraorbital areas of women with dark circles compared with

those of women without dark circles. In addition, we were able

to detect improvement in the dark circles after 6 weeks’ use of

anti-dark circle cosmetic products by visualizing the distribution

of the relative concentrations of melanin and hemoglobin oxy-

gen saturation ratio.

Conclusion: Using a spectral camera, we developed a non-

contact image-processing system that was capable of captur-

ing a wide area of the face to visualize the distribution of the

relative concentrations of skin chromophores in the face.

Key words: dark circles – hemoglobin oxygen saturation ratio

– melanin – spectral imaging system

� 2013 John Wiley & Sons A/S. Published by JohnWiley & Sons LtdAccepted for publication 3 May 2013

MELANIN AND hemoglobin are the dominantchromophores in human skin. Melanin is

a black or dark brown chromophore of the epi-dermis. The primary determinant of variabilityin skin color is the amount, density, and distri-bution of the pigment melanin. Hemoglobin,which is the primary protein constituent of redblood cells, is a complex molecule responsiblefor transport of oxygen throughout the body.Oxygenated hemoglobin (oxyhemoglobin) has areddish hue and produces a pinkish tint inlightly pigmented skin. Deoxygenated hemoglo-bin (deoxyhemoglobin) has a purplish color andproduces the bluish tint in lightly pigmentedskin that is characteristic of oxygen deprivationand suffocation.Dark circles or infraorbital dark circles, a skin

condition in which there is relative darkness ofthe infraorbital eyelids, is a major cosmeticproblem. It has been reported that dark circles

are caused by a process of dermal pigmentationand/or hemodynamic congestion (1–4). Con-cerning blood congestion in the lower eyelids,Matsumoto et al. (1) reported that the amountof blood in the lower eyelids is larger and thevelocity of blood flow is slower than in thecheek. They concluded that dark circles appearin areas where there is stasis or congestion ofcutaneous blood. Masuda et al. (2) investigateddark circles in Japanese women using a spectro-photometer, and found increased amounts ofmelanin and hemoglobin, decreased hemoglo-bin oxygen saturation ratio, and slow bloodflow in the dark circle areas.In most previous studies, dark circles in the

lower eyelids have been evaluated using con-tact-type spectrophotometers to determine thecolor value, and melanin and hemoglobin con-tents. Although this instrument can measurethe target site with high accuracy, it is difficult

1

Skin Research and Technology 2013; 0: 1–9Printed in Singapore � All rights reserveddoi: 10.1111/srt.12074

© 2013 John Wiley & Sons A/S.Published by John Wiley & Sons Ltd

Skin Research and Technology

to obtain these measurements in the dark circleareas because of the small aperture of the probehead. Ohshima and Takiwaki (4) proposed animage-processing method for analyzing andquantifying dark circles. They determined theerythema index, melanin index and oxygenationindex from an original digital RGB color imageof the eyelids, and visualized their distributionin the face. Evaluation of dark circles using ery-thema index and melanin index images seemsmore suitable than the use of a contact-typereflectance meter. However, complete separa-tion of erythema and melanin has not yet beenachieved.Recently, the spectral camera has been widely

used in many fields, for example in medicineand food science. The spectral camera is animaging spectrometer that provides full andcontiguous spectral information for each pixel.By analyzing spectral data in each pixel of thespectral image, it is possible to construct distri-bution maps of the relative concentrations ofmelanin and oxy- and deoxyhemoglobin.In this study, we developed a spectral imag-

ing system that was capable of capturing awide area of the face to quantify and visualizethe distribution of the relative concentrations ofskin chromophores including melanin, and oxy-and deoxyhemoglobin. We then examined thecharacteristics of dark circles using this systemand visualized the distribution of the relativeconcentrations of melanin and hemoglobin oxy-gen saturation ratio, which is the ratio of oxyhe-moglobin to total hemoglobin. We alsoevaluated the efficacy of an anti-dark circle cos-metic by visualizing changes in the local distri-bution of the skin chromophores.

Materials and Methods

Spectral imaging system for facial skinThe original spectral imaging system for facialskin consisted of a spectral camera and an illu-mination unit (Fig. 1). The spectral camera(Hyper Spectrum Camera, HSC1700; HokkaidoSatellite Corporation, Tokyo, Japan) wasequipped with a transmission grating and anarray sensor with an eight-bit monochromeCCD camera with 640 9 480 pixels, and had aspectral range of 350–1050 nm containing 141bands with 5-nm resolution. The opticalunit consisted of both a spectrometer and ascanning mechanism using an internal digital

servomotor. The camera was capable of takinga full-sized spectral image every 16 s. The dis-tance from the camera head to the subject’s facewas 460 mm. The illumination unit (paintedinside matte white) was designed to providediffuse illumination of the subject’s face to elim-inate artifacts from specular reflections. TwelveLED lamps (model EXSBN; CCS Ltd., Tokyo,Japan) were utilized as the light source, whichemit an almost constant reflectance in the wave-length range 400–800 nm. A warm-up periodwas included to stabilize the light intensity. Inaddition, the positioning device was placed onthe imaging system to align the subject’s faceduring image acquisition.

Analysis of the spectral imageFigure 2 shows the data analysis algorithms.The spectral images obtained with the spectralcamera were formatted using the band inter-leaved by line (BIL) format. The image was thennormalized as spectral reflectance data format-ted in BIL using a reference white image (WhiteCalibration Plate No. 9; Murakami ColorResearch Laboratory, Tokyo, Japan). Spectralreflectance data for each pixel were convertedinto apparent absorbance (A = log10 (1/reflec-tance)). The multiple regression analysis wasperformed using the absorbance values in thewavelength range 500–700 nm at 10-nm inter-vals to calculate the melanin, oxyhemoglobin,and deoxyhemoglobin contents (5). In addition,hemoglobin oxygen saturation ratio was calcu-lated from the ratio of oxyhemoglobin to totalhemoglobin.

Fig. 1. Spectral imaging system for facial skin. The system consistsof a light source, a spectral camera, and a computer.

2

Kikuchi et al.

Morphing technology for spectral imageTo analyze group mean data, a face morphingalgorithm for spectral images was proposed.Given several spectral images of subjects, wecreated a spectral image of the face by morp-hing based on all 41 control points. The controlpoints consisted of 19 facial feature points,including the end points of the eyes, the mouth,the nose, and the eyebrows. In addition, wegenerated 22 more feature points around theface edge, which were the points of intersectionof the extension lines of the first 19 facial fea-ture points with the face edges. A triangulationmethod was then used to match and warp thesource image to the target image using the con-trol points. This morphing technology was per-formed pixel by pixel, and each of the 31spectral reflectance values in the wavelengthrange 500–700 nm at 10-nm intervals was dealtwith individually, instead of each of the colorcomponents RGB used in conventional colormorphing.

Validation of the spectral imaging systemTo validate the system, the spectral reflectanceof the cheek and lower eyelid of 32 Japanesewomen (age range 34–59 years; mean age44.3 years) was measured using a contact-type

spectrophotometer (CM700d; Konica MinoltaSensing, Tokyo, Japan) with a probe head aper-ture of approximately 7 mm at the objectivearea. At the same time, spectral reflectanceimages of the whole face of the same 32 subjectswere obtained using the system. In all spectralreflectance images of the face, regions of inter-est were selected in the same area as measuredusing the contact-type spectrophotometer.Melanin, oxyhemoglobin and deoxyhemoglobinand Laboratory values in the CIE1976 L*, a*, b*color space (CIELAB) were calculated fromspectral reflectance, and the correlationsbetween the two types of measurement systemwere determined.

Relationship between relative concentrations of skinchromophores and visual evaluation of dark circlesThe relationship between the distribution of therelative concentrations of skin chromophoresand visual evaluation in terms of the intensityof the dark circles was determined. Fifty-ninehealthy Caucasian female volunteers (age range30–59 years; mean age 39.2 years) with or with-out dark circles were enrolled. Of these, 30 sub-jects had been conscious of dark circles in theirappearance, while the others had not. Theintensity of the dark circles was evaluated

Fig. 2. Schematic diagram of the skin color analysis technique. Spectral data from each pixel in the image are analyzed and separated intomelanin, hemoglobin value, and hemoglobin oxygen saturation ratio.

3

Imaging of hemoglobin oxygen saturation ratio

visually by three cosmetic investigators using ascale of 1–3 (3 marked, 2 moderate, and 1 nor-mal). The intensity of the dark circles was eval-uated using both a conventional contact-typespectrophotometer and the original spectralimaging system.To compare the distribution of the relative

concentrations of the skin chromophores in theface between the groups with dark circles andwithout dark circles, each of the average spec-tral images was constructed from the spectralimages from eight women using a morphingprocedure. The distribution of the relative con-centrations of melanin, hemoglobin, and hemo-globin oxygen saturation ratio in the face wasthen determined. All measurements were per-formed under controlled environmental condi-tions (temperature 22°C, relative humidity45%).

Evaluation of efficacy of the anti-dark circle cosmeticTwenty-two healthy women (age range24–56 years; mean age 40.0 years; 7 Asian, 10Hispanic, and 5 Caucasian) with dark circleswere enrolled in this study. Volunteers hadmoderate to marked dark circles, and did nottake any medication during the study. The anti-dark circle cosmetic was designed to reduceblood congestion and melanin, and was appliedtopically twice a day (morning and evening) tothe infraorbital area for 6 weeks. The spectralreflectance of the lower eyelids and cheek wasmeasured using a spectrophotometer undercontrolled environmental conditions (tempera-ture 22°C, relative humidity 45%) and spectralimages around the eyes were obtained at 2, 4,

and 6 weeks. The study protocol was approvedby the Ethics Committee at Shiseido ResearchCenter. All subjects were informed about theconditions of the study and written informedconsent was obtained in all cases.

Statistical analysisData are expressed as means � SD unlessotherwise indicated. The linear correlation coef-ficient was used to evaluate the accuracy of thesystem. The unpaired two-tailed Student’s t-testwas used to evaluate the condition of the darkcircles and Student’s t-test for paired samples.All tests were two-sided, and P < 0.05 wastaken as the significance level to evaluate theeffectiveness of the skin-care product.

Results

Validation of the spectral imaging systemFigures 3 and 4 show comparisons between thespectrophotometer and spectral images for thecheek and lower eyelids, respectively. The cor-relation coefficients between the spectropho-tometer and spectral images were higher than0.900 for all parameters for both the cheek andlower eyelids. These results confirmed that oursystem for facial imaging was sensitive enoughto offer analyses of melanin, oxyhemoglobin,and deoxyhemoglobin with a good spatialdistribution.

Characterization of the dark circlesFigures 5 and 6 show the results of themeasurement at the lower eyelids with the

(a) (b) (c)

Fig. 3. Comparison between the facial spectral imaging system and spectrophotometer for the evaluation of melanin content (a), oxyhemoglobin(b), and deoxyhemoglobin (c) in the cheek. The solid lines are regression lines calculated by the method of least squares, and R is the correlationcoefficient.

4

Kikuchi et al.

(a) (b) (c)

Fig. 4. Comparison between the facial spectral imaging system and spectrophotometer for the evaluation of melanin content (a), oxyhemoglobin(b), and deoxyhemoglobin (c) in the lower eyelid. The solid lines are regression lines calculated by the method of least squares, and R is the correla-tion coefficient.

(a) (c)(b)

Fig. 5. Melanin content (a), hemoglobin oxygen saturation ratio (b), and hemoglobin content (c) of the lower eyelids in Caucasian women with/without dark circles measured by the spectophotometer. *P < 0.05, ***P < 0.001.

(a) (c)(b)

Fig. 6. L* (a), a* (b), and b* (c) values of the lower eyelids in Caucasian women with/without dark circles measured by the spectrophotometer.*P < 0.05, **P < 0.01, ***P < 0.001.

5

Imaging of hemoglobin oxygen saturation ratio

spectrophotometer. Mean values of melanin,hemoglobin oxygen saturation ratio, and hemo-globin were calculated in each group (15 womenwith marked dark circles, 10 with moderate, and34 without as controls). The mean value of mela-nin in the lower eyelids was significantly higherin women with dark circles than in the controls(Fig. 5a). The mean value of hemoglobin oxygensaturation ratio in the lower eyelids was signifi-cantly lower in women with dark circles than incontrols (Fig. 5b). On the other hand, no signifi-cant differences in hemoglobin were foundamong the three groups (Fig. 5c).The mean values of L *, a *, b * were then cal-

culated from the spectral reflectance. The skinlightness (L* value) in the lower eyelids of thewomen with dark circles was significantlylower than in women without dark circles(Fig. 6a). Conversely, the yellowness parameter(b* value) in the lower eyelids was significantlyhigher in women with dark circles than in con-trols (Fig. 6c). Average spectral images ofwomen with and without dark circles areshown in Fig. 7. The amounts of melaninincreased and hemoglobin oxygen saturationratio decreased locally in the infraorbital areasof women with dark circles compared with theamounts in women without dark circles.

Improvement in dark circles with the anti-dark circlecosmeticFigure 8 shows the changes in the amount ofmelanin, hemoglobin oxygen saturation ratio

and the amount of hemoglobin measured usingthe spectrophotometer in the lower eyelids andcheek during 6 weeks of treatment. The amountof melanin in the lower eyelids decreased grad-ually, but did not change in the cheek (Fig. 8a).The hemoglobin oxygen saturation ratio in thelower eyelids increased significantly over2 weeks and then remained constant up to6 weeks, while the levels in the cheek did notchange (Fig. 8b). The amount of hemoglobin inthe lower eyelids decreased significantly over6 weeks, and a similar trend was seen for thecheek (Fig. 8c).Figure 9 shows the changes in L*, a*, b* mea-

sured using the spectrophotometer. Applicationof the anti-dark circle cosmetic for 2 weeks ledto a statistically significant increase in L* valueof the lower eyelids (Fig. 9a). The rednessparameter (a* value) remained unchanged inthe lower eyelids and the cheeks (Fig. 9b). Thevalue of b* in the lower eyelids decreased up to4 weeks (Fig. 9c). To visualize the changes indark circles following continuous use of theanti-dark circle cosmetic, distribution maps ofthe relative concentrations of skin chromo-phores were created from the spectral images.The spectral reflectance images of the face weremorphed into those of an averaged Asian, His-panic, and Caucasian female face, each of whichwas constructed from the images from fourwomen. After 6 weeks of use, the amount ofmelanin in the infraorbital area decreased andan improvement in pigmentation was observed(Fig. 10a). Hemoglobin oxygen saturation ratio

(a)

(b)

(c)

Fig. 7. Average spectral images of women with dark circles (left) and without dark circles (right). (a) Original images obtained using the facialspectral imaging system, (b) melanin distribution images, and (c) hemoglobin oxygen saturation ratio images.

6

Kikuchi et al.

increased after 2 weeks of use, and the micro-circulation was improved (Fig. 10b).

Discussion

In this study, we established a non-contact sen-sitive imaging system, consisting of a spectralcamera and an illuminating light unit, for mea-suring distribution of skin chromophores in theface with high reproducibility. We applied this

system to the characterization of dark circlesaround the eye and evaluation of efficacy of ananti-dark circle cosmetic, and successfullyrevealed an increase in melanin content and adecrease in hemoglobin oxygen saturation ratioin dark circles and their improvement by serialuse of anti-dark circle cosmetic, not only asnumerical data, but also as comprehensivevisual presentations around the eyes created byspectral morphing technology. Recently, several

(a) (c)(b)

Fig. 8. Changes in the mean values of melanin content (a), hemoglobin oxygen saturation ratio (b), and hemoglobin content (c) measured usingthe spectral photometer in the lower eyelids (closed circles) and the cheek (open triangles) during 6 weeks of cosmetic treatment. +P < 0.10,*P < 0.05, **P < 0.01.

(a) (c)(b)

Fig. 9. Change in mean values of L* (a), a* (b), and b* (c) measured using the spectrophotometer in the lower eyelids (closed circles) and the cheek(open triangles) during 6 weeks of cosmetic treatment. +P < 0.10, **P < 0.01, ***P < 0.001.

7

Imaging of hemoglobin oxygen saturation ratio

authors have used a spectral camera for theassessment and visualization of skin chromo-phores (6–9). They evaluated skin lesions suchas port-wine stains, angioma, nevi, and mela-noma using spectral images in the visible andinfrared regions (i.e. 400–970 nm) obtained withliquid crystal tunable filters or a series of dis-crete band-pass filters. However, their targetarea was limited to a narrow area on the fore-arm, and the target skin chromophore was spe-cific to only one type. Thus, a system that canvisualize the distribution of all components ofskin color including the amounts of melaninand hemoglobin, and hemoglobin oxygen satu-ration ratio in a wide area of the face has notyet been developed.In contrast, our newly developed spectral

imaging system consisted of a spectral camerawith gratings and prisms, and a lighting unitfor uniform irradiation of the face. We werethen able to construct a distribution map ofthe relative concentrations of skin chromo-phores in a wide area of the face by multipleregression analysis of spectral data (5). Incomparison with the data obtained using amulti-band camera equipped with color filters,more detailed spectral data are acquired usingour system.Caucasians with dark circles showed larger

amounts of melanin and lower hemoglobin

oxygen saturation ratio and L* values in thelower eyelids than subjects without dark circles(Figs 5 and 7). From these results, it is consid-ered that pigmentation and blood congestion inthe skin induce decreases in L* values of darkcircles. Regarding the pathogenesis of pigmen-tation and blood congestion around the eyes,there are a few published studies (10, 11). Thesestudies showed that pigmentation is caused bycongenital and environmental factors. The con-genital factors include dermal melanin deposi-tion seen on histology and several types ofbenign pigmented lesions. The environmentalfactors include excessive sun exposure, drugingestion, and postinflammatory pigmentationsecondary to atopic or allergic contact dermati-tis. Blood congestion is caused by anatomic fac-tors, which include superficial locationvasculature and the visible prominence of thesubcutaneous vascular plexus. However, theprecise cause has not been clearly elucidated,and ethnic differences in the skin structure ofthe infraorbital area and the mechanisminvolved in the formation of dark circles stillhave not been clarified.In a previous study in Japanese women (2),

melanin accumulation and blood congestionwere considered factors that could induce darkcircles in Caucasians. Conversely, in this studythe results on hemoglobin were different from

(a)

(b)

Fig. 10. Mean spectral images of (a) melanin distribution and (b) hemoglobin oxygen saturation ratio of dark circles before and after cosmetictreatment.

8

Kikuchi et al.

those in the previous study (2). Japanesewomen with dark circles showed largeramounts of hemoglobin in the lower eyelidsthan those without dark circles, when measur-ing skin chromophores using a portable reflec-tance spectrophotometer. In contrast, noobvious difference was observed in this studyin Caucasians (Fig. 5c).Regarding the use of the anti-dark circle cos-

metic, the changes were not seen in untreatedcheek areas. In contrast, the hemoglobin oxygensaturation ratio increased, and the melanindecreased in the infraorbital areas treated withthe cosmetic (Figs 8 and 10). The increases inhemoglobin oxygen saturation ratio were con-sidered to be caused partly by massaging thecream around the eyes, and the effects on themelanin content might have been mediated, atleast in part, by the vitamin C in the cream. Alarge trial is necessary to investigate the effectsof racial differences on the mechanism involvedin the generation of dark circles, and to investi-

gate changes in the orbital skin after continuoususe of cosmetics.

Conclusion

The use of a spectral imaging system whichwas able to detect the precise distribution of therelative concentrations of skin chromophoresshowed local increases in melanin content anddecreases in hemoglobin oxygen saturation ratioin the infraorbital areas of women with darkcircles. In addition, we confirmed that dark cir-cles could be improved by using an anti-darkcircle cosmetic as evaluated by visualizing thedistribution of melanin concentration andhemoglobin oxygen saturation ratio, which arethe factors that induce dark circles.

Acknowledgement

The authors thank Dr Mariko Egawa for herhelpful discussion and suggestions.

References

1. Matsumoto M, Kobayashi N,Osamu H, Arai S. Study on themechanisms associated with darkcircles. J Soc Cosmet Chem Jpn2000; 34: 152–159.

2. Masuda Y, Takahashi M, Satou A,Yanai M, Yamashita T, Iikura T,Ochiai N, Ogawa K, Sayama K. Der-matological study on dark eye cir-cles and their treatment with newlydeveloped cosmetics. J Soc CosmetChem Jpn 2004; 38: 202–210.

3. Fukuda Y, Soga H, Satoh H, KitahataT, Yoshizuka N, Takema Y. Spectro-scopic characterization of color poly-morphism in the orbital skin. J SocCosmet Chem Jpn 2005; 39: 195–200.

4. Ohshima H, Takiwaki H. Evalua-tion of dark circles of the lowereyelid: comparison between reflec-tance meters and image processingand involvement of dermal thick-

ness in appearance. Skin Res Tech2008; 14: 135–141.

5. Masuda Y, Yamashita T, Hirao T,Takahashi M. An innovativemethod to measure skin pigmenta-tion. Skin Res Tech 2009; 15: 224–229.

6. Balas C, Themelis G, Papadakis A.A novel-spectral imaging system:application on in-vivo detectionand grading of cervical precancersand of pigmented skin lesions. InProc of “Computer Vision Beyondthe Visible Spectrum” CVBVS’01Workshop, Hawaii, USA, Dec.2001.

7. Randeberg LL, Winnem AM,Langlois NE. Skin (Los Angeles)changes following minor trauma.Laser Surg Med 2007; 39: 403–413.

8. Jakovels D, Spigulis J. 2-D map-ping of skin chromophores in thespectral range 500–700 nm. J Bio-photonics 2010; 3: 125–129.

9. Kobayashi M, Ito Y, Sakauchi N,Oda I, Konishi I, Tsunazawa Y.Analysis of nonlinear relation forskin hemoglobin imaging. OptExpress 2001; 9: 802–812.

10. Freitag FM, Cestari TF. What causesdark circles under the eyes? J Cos-met Dermatol 2007; 6: 211–215.

11. Roh MR, Chung KY. Infraorbitaldark circles: definition, causes andtreatment options. Dermatol Surg2009; 35: 1163–1171.

Address:Kumiko KikuchiShiseido Research Center2-11-1, HayabuchiTsuzuki-kuYokohama 224-8558JapanTel: +81-45-590-6000Fax: +81-45-590-6387e-mail: kumiko.kikuchi1@to.shiseido.co.jp

9

Imaging of hemoglobin oxygen saturation ratio