Developing a custom calibration target for objective skin color measurement with color corrected dermoscopy

Summary: Abstract Body: Reliance on subjective skin color assessment can perpetuate inequities for patients with skin of color(SoC). Erythema appears more subtly in SoC, making it difficult to assess/monitor. Subjective assessment also limits diversity in clinical trials and AI image training sets. Colorimeters are used to objectively assess skin color in the L*a*b* color space or individual typology angle(ITA=atan[(L*-50)/b*]) to estimate melanin but can be costly or bulky for curved/small anatomic sites. Photography can be an accessible colorimetry tool but is limited in color consistency due to non-uniform illumination and camera calibration. We hypothesized a dermatoscope can be repurposed as a colorimeter when combined with a color calibration target. A pilot study with 23 subjects with dark to light pigmentation(ITA-52 to 65) tested the ability of color-calibrated dermoscopy to estimate ITA using a 9.5x5.6mm standard calibration target with 30 colors. Spectrophotometry was performed at each site to measure gold-standard L*a*b* values and calculate ITAGS. For each image, a correction matrix was applied to minimize the color difference between known and measured L*a*b* values in the target. L*a*b* values were extracted from the uncorrected and corrected images to calculate ITA(ITAcor, ITAuncor). ITAcor had lower mean error(ME=18°±9.8) and stronger correlation(r=0.96) than ITAuncor(ME=29°±24,r=0.77) when compared to ITAGS. However, standard targets occupy most of the dermatoscope’s field of view(FOV), and their colors do not represent skin tones. To improve calibration for skin color while integrating colorimetry into dermatoscopes, a 21mm diameter custom target with 24 patches(each 1.6x1.6mm), was designed and 3D printed to fit peripherally in the dermatoscope’s FOV. The target is made from the Pantone SkinTone Guide with known L*a*b* values from diverse skin reflectance spectra. Preliminary results show a more accurate estimate of L*a*b* values. The development of this target suggests that color-corrected dermoscopy can provide easily accessible and objective skin color assessment. Maysoon Harunani¹, Anmol Jarang¹, Leonid Shmuylovich¹ 1. Washington University in St Louis, St. Louis, MO, United States. Bioinformatics, Computational Biology, and Imaging