Exploring sensitive skin to design reliable measurements

Dear Editor, Sensitive skin (SS) is a sensory syndrome in which stimuli usually not inducing reaction lead to unpleasant sensations (burning, pain, pruritus, tingling, etc.).1 Generally associated with any clinical sign, SS can be triggered by chemical, environmental or physiological factors.2 Objective assessment of severity remains challenging because its subjective nature and the lack visible symptom. To fill that gap, this letter proposes two methods quantitatively assess severity. A evaluation was conducted during winter 2019 90 Caucasian females (1865 years, mean ± SD = 42 14, phototype I–IV). The subjects answered 60 yes/no questions self-assess SS-inducing factors reactions 4 estimate their sensation intensities (none, slight, moderate, marked). Based on these four last replies, trained assessor classified as non-sensitive sensitive. lactic acid sting test3 also provided stinger/non-stinger classification. SpectraFace hyperspectral front face images4 all were collected. In questionnaires, tended select cheeks cheekbones most sensitive facial region (34% 28% respectively) against mouth contour, nose, forehead, eye contour. mapping frequency scores presented Figure 1. Comparison distribution basic characteristics (typology, shine, redness, dryness) between panel (70 assessor) one (20 subjects) revealed no difference. Even if are combined linear regression, classification prediction reliable: Matthew's correlation coefficient5 (MCC) equals 0.49, good precision (0.95) poor (0.45). Most declared stinger-negative (Figure 2), confirming test an imperfect predictive tool.6 better apprehend subtle perception over using existing data, regression analysis generate continuous scale instead binary inputs questionnaire answers, turned into numerical (No 0, Yes 1, None Slight Moderate 2, Marked 3) targets (Non-sensitive 1). For readability reasons, output score rescaled 0 (not sensitive) 10 (very sensitive), leading score. By thresholding limit set 2.3, defined tests trials, expert categorisation could then predicted MCC 0.97, 0.98 1.00 (Table Therefore, based questionnaires improved description self-assessed sensitivity. However, it requires rather long questionnaire. ensure unbiased must asked guaranteeing comprehension. strengthen approach measurement, decided rely objective data from images endpoint more robust precise tool, machine-learning [ML] model. ML model would generalise information group subjects, making small individual imprecisions self-assessment so impacting. From images, colour parameters (CIE L*a*b*, CIE L*C*h, ITA, IWA7), homogeneity (H76), oxygen saturation, melanin haemoglobin concentrations computed, several zones (cheeks, cheek bones, nasolabial folds). multi-layer perceptron [MLP] subset 80% randomly selected panel, above-described ground truth. coefficient determination R2 0.93 reached training 0.81 4) remaining 20% data. instrumental index, called Sym'Index, reliably sole acquisition subject's face. experiment index ability follow evolution, another double-blind 2020 25 previous panel. Two products compared: soothing formula (VIVO 1902.A.02) placebo 1902.B.02). contains SymRelief 100 (INCI name: Bisabolol, Zingiber Officinale (Ginger) Root Extract). It synergistic blend synthetic bisabolol high purity natural ginger extract. This ingredient offers comprehensive solution targeting inflammatory mediators (IL-1α & PGE-2)8. acquired just before product application [D0], after 5 days [D5] [D10] repeated applications (twice day [morning/evening] clean dry skin, much needed until complete absorption), analysed. indices automatically computed for each half-face at time point. Results 2) indicate that, overtime, tends induce decrease while increase index. differences significant, changes induced remain coherent promising. impacted enough slight aspect highlight effects. conclude, initial opens new perspectives assessment, first imaging artificial intelligence combining classical appraisal innovative, instrumental, multiparametric analysis, highly correlating expert's assessments. proposed methodology help up sensitivity evolution evaluate effects products. next steps, will underline finer evolutions, increasing size grades, truth repeatable (asking times questionnaire), adding other areas removing redundant parameters. support findings study available corresponding author upon reasonable request.