Virtual approach of the aesthetical fit between hair colours and skin tones in women of different ethnical origin backgrounds

To determine the aesthetical accordance between a given skin tone and the 11 possible colours of head hairs, covered by a marketed hair colouration product.


INTRODUCTION
Despite their common genome, humans express a wide diversity of phenotypes that comprise different sizes, builds, weights, shapes and colour of hairs, face or eyes. . . In short, 'all the same, all different' (A. Langaney).
Two phenotypical elements are important drivers of the human appearance, skin and hair, particularly through a common denominator: their innate colour mainly conveyed by melanin pigments, synthetized by their own underlying melanocytes. Once mature, these pigments (eumelanin/black and phaeomelanin/red) are further injected as tiny granules either within epidermal cells or into the hair cortex. All these events and their consequences have been the objects of an abundant literature. [1][2][3][4][5][6][7][8][9][10][11][12][13][14] As for a healthy skin, these processes lead to a constitutive palette of tones, from pale (Keltic) to dark (African, Indian), due to an increased amount of both melanin pigments, noting that in all cases the ratio eumelanin/phaeomelanin remains constant (76% vs. 24%, respectively). 15 Using spectrometers, the range of skin tones is now described by the continuous Individual Typology Angle parameter (ITA), calculated from the L*a*b* system (ITA • = arctan [L* − 50 / b*] * 180 / π), from −10 • up to > +55 • from very dark to pale complexions. 16,17 As skin tone is prone at changes with sun exposures (tanning), ITA allows to record either the intensity of a skin darkening, to evaluate the photo-protective effect of sunscreens or to quantify a possible whitening effect. [18][19][20][21][22][23][24][25][26][27] Skin tone and its homogeneity may vary, too, with age (dark spots, i.e., lentigines) or with some discolouring afflictions such as vitiligo (whitening) or melasma (darkening). It is in addition a social/aesthetical element of high importance, especially in Asian countries, where a darker skin is negatively perceived, as a signature of 'low-class' people (daily exposed to sun such as farmers, fishermen, outdoor workers etc.), realizing the high demand of topical whitening products in these countries. The most ancient procedure used to slightly correct the innate skin tone, and its possible heterogeneity is the technique of making-up, based on powders of different shades applied by gentle brushes. Make-up is nevertheless of a very provisory aspect and should be adapted to all skin tones, worldwide. 28,29 It allows to propose a first clusterization of skin tone, leading to 15 sub-groups of bare skin colour to represent different ethnic groups. Another recent approach, 30 combining colour science, consumer and colour artists insights allowed to cocreate the ideal shade range through a strategy analysis based on the identification of six global skin tone clusters with 18 sub-clusters across the world. All these approaches lead to a dedicated and specific clusterization of skin tone but carried out on the same population diversity.
As for hair colour or hair tone (HT), the status appears somewhat different, with similarities and divergences. Unlike skin, hair colour tends to fade with sun or ultraviolet (UV) exposures via oxidative processes. [31][32][33] Aging leads to a progressive loss of melanocytes of the hair follicle, 34,35 ultimately leading to hairs fully deprived of pigments, perceived as white hairs, a dreaded signature of aging. Of note, on a worldwide aspect, it comes clear that brown to dark hairs largely predominate, even in non-Hispanic European and Euro-American population who present the widest palette of hair colours, red hairs included. 36,37 But a striking divergence with skin emerged, about a century ago, by the development of oxidative hair dying using, as example, Para-PhenyleDiamine that, in the presence of other reactive entities (couplers) under oxidative conditions, leads to various hair colours. 32 It is a safe procedure daily performed in hair salons by millions of women, worldwide. As for men, many aged men with brown/dark hairs resort to oxidative hair dying to cover their undesired white hairs.
In brief, the human faces, on a worldwide vision, present a large mosaic of variations created by different hair and skin colours, ranging from the darkest to the palest tones. Would an 'ideal' aesthetical fit between skin and hair colours be found (or not) across different cultural/ethnical criteria?, was the driving question of the present work.
Taking into account that skin tones not only vary between ethnical ancestries (e.g., African American vs. East Asian or non-Hispanic European) but show differences within the same large ethnical group, we therefore aimed at evaluating by a virtual process, based on an artificial intelligence (AI)-based algorithm, the aesthetical fit between some skin tones and some hair colours artificially modified by a marketed hair dyeing product. Such virtual approach was chosen as applying, in vivo, 11 different hair colouring on the models was unfeasible for both ethical and technical reasons. Same limitation obviously holds true with regard to skin innate colour. Hence, when virtually applied on seven professional top models representing four ethnical ancestries, changes in skin and hair tints were obtained:

Subjects
The images of the top models illustrated here correspond to women under contract with our Group that covers the right to imaging. The

Data basis
The present study initially re-collected and used data on skin and HTs 19,36 previously obtained and published by our group of research.
The first ones dealt with the natural skin tones of 507 women of ethnical backgrounds, objectively defined through the L*a*b* reference scale, using the Chromasphere methodology, 19 thus allowing to create a skin colour chart. 19 Similar to such objectives, the second study analyzed, through a spectrophotometer, the natural colour of 2870 hairs sampled from 2057 subjects (both genders) of 23 different regions of the five continents. 36 In all cases, the L* (Luminance), a* (red-green vector), and b* (yellowblue vector) values were used as such or as components of other colour markers such as hue (arctan b*/a*) or chroma (√(a* 2 + b* 2 ). The latter work allowed to better define 10 different HTs fitting well with the scale used by hair professionals dealing with hair dyeing, where HT1 represents the blackest tone, up to HT10, the palest one (extreme blond).

Protocol for cluster definition
Skin clusterization: L*a*bĺ* references of skin tones were re-extracted from previous data basis, 19,28    Step 2 dealt with colour assessment, where all pictures were presented one by one in a randomized order. The panel had to answer to the two types of questionnaires through 3 simple wordings, that is, agree, disagree, and do not know. The two series of questions are shown in Table 3.

Statistics/Handling of data
where p is the observed percentage of the sample, n its size, t a multiplier at a confidence level (here α = 95%), and I is the maximum of p  Table 4  Euro-American cohorts present slightly lower positive answers than the three other groups.

RESULTS
In Step 2, the independent evaluation of each pictures permitted to go deeply in the specific analysis of the appreciation between skin tone and hair colour. Compiling all these answers with their respective modified HTs allows to precise the HTs that were given privilege by the five TA B L E 2 Repartition of the constitutive hair colours of the assessing panels. Numbers indicate the average hair tone (HT) established by our hair dying experts during the enrollment phase

TA B L E 3
The two questionnaires (translated in four languages, English, Chinese, and French) used in the two-step procedure of assessments Step were unable to clearly associate their positive answers with the wordings used (young, healthy etc.). Despite these lower values than those TA B L E 4 Summary of the positive assessments ('agree' as %) obtained by the five local panels on the aesthetical adequation between hair colours according three skin tones on Step 1 (whole hair colour range)

TA B L E 5 Repartition of modified hair tones (HTs) vis à vis their global appreciations by the local panels, according to origin and skin tones on
Step 2 (evaluation on single picture) TA B L E 6 Summary of the positive answers ('agree' , as %) obtained by the five local panels on various subjective social aspects of the facial appearance of virtual models of the three skin tones, where hair colours were artificially modified (obtain on Step 1, on the whole range of hair colour) Would you say that, wearing this hair colour, this woman looks. . .

DISCUSSION
The human face presents, worldwide, a mosaic of variations in the intrinsic colours of skin and hairs, keeping in mind that innate darker skin complexions and darker hairs largely predominate. Whatsoever, in their constitutive status, skin and hair colours were shown potent drivers of social or psychosocial relationships, [38][39][40][41] although these quoted works were performed on women of same ethnical ancestry or on lightly pigmented young women.
As hair dying became, since decades, a common practice in many women, the paradigm of these quoted works may have changed.
Such a question mostly led us to undertake the present study based on virtual images that included a mix of skin tones and artificially modified hair colours in women from five ethnic ancestries. In brief, we believed that the virtual modifications of hair colour and skin tones would largely overcome such issues. Of note, the present work only dealt with aesthetical assessments and associated subjective terms (young, healthy, glowing. . . ) and not on their possible social or psychosocial consequences that were only explored on innate skin and HTs. [38][39][40][41] The results shown in Table 5 (4)(5)(6). Globally, the divergences between the assessments by the non-Hispanic European US and French panels remain rather coherent despite a few discrepancies, again likely linked to current cultural and fashion criteria noting that the skin tone (L, M, D) has nevertheless a rather minor impact upon the assessments of HTs. We acknowledge that the present work obviously possesses limits with regard to the statistical aspect of the answers from all panels who classify many HTs with probably too much indulgence, leading to many differences of low amplitudes between HTs. Hence, these answers should be more regarded as trends, keeping in mind that these panels may not fully reflect the assessments of their whole population. In addition some terms, such as glow, used by laymen are probably too subjective, leaving room to uncertainties. Whatsoever, the methodological approach by a virtual process has proven interest by using aesthetical models on which two major phenotypes (skin and hair colours) can be infinitely modified. In addition, such approach may allow to confidentially help subjects with skin problems or afflictions (melasma, albinism, vitiligo, lentigines, freckles, and telangiectasia etc.) to help them choosing the more appropriate hair colouring tone. This method appears as a good tool to create material for test when it is not possible to assess the desired rendering. However, the final rendering is always based, and finetuned on real tests carried out on models, to make sure that the simulation globally fits with reality. Mixing skin tone and hair colour on the same person allows to propose a standard experience of evaluation to consumer in limiting the influence of facial differences that could induce preferences not only linked to the association between hair and skin colours. Using different representative ethnical origin models, we planned a worldwide study that may allow to show the diversity of consumer preferences, linked to cultural aspects and trends in each country. Even if this study is today conducted in CRO, to control the picture visualization on a calibrated screen, future studies will be completely performed in a full digital way, that is, through the nomad selfacquisition (selfie), characterizing skin and hair features by gathering a AI-based algorithm 42 and a nomad consumer study. Despite some inherent limits, the present work appears at the crossroad between genetic backgrounds and the secular and/or cultural desire to improve and amend their inherited entailments at an intimate aesthetical level.

ACKNOWLEDGEMENT
All costings of the study were entirely met by the L'Oréal Research and Innovation Department.