701 Rethinking alopecia scores: Correlating measures of hair loss with established photographic scoring systems

Abstract

Alopecia scoring systems have traditionally been designed to assess and track severity and progression of specific types of alopecia. Each scoring system’s design and clinical validation is limited to only one alopecia type (e.g., Severity of Alopecia Tool score for alopecia areata (AA), Sinclair visual analogue scale for female pattern hair loss (FPHL), Olsen top extent scale for central centrifugal cicatricial alopecia (CCCA)), even though all forms of alopecia share a common feature of decreasing hair density. We propose that a scoring system that measures percent hair loss would be useful across different types of hair loss. We tested correlations between the AA, FPHL, and CCCA scoring systems and the underlying percent hair loss. On a dataset of 284 top-view scalp images taken from 250 subjects, 3 raters manually scored each image using the 3 different scoring systems. After an intraclass correlation analysis demonstrated reliability between the raters, we established a final consensus set of 3 scores per image. Using regression models, we analyzed the score correlations within and across alopecia types. Our results suggest that there is a precise and quantifiable correlation between these scoring systems and the underlying hair loss percent: both CCCA and FPHL scoring systems show a logarithmic relationship with respect to percent scalp area affected (a proxy for the AA scale), resulting in an R2 of 0.793 and 0.804, respectively. Additionally, the CCCA and FPHL scoring systems show a linear relationship with each other, resulting in an R2 of 0.963. These findings suggest that distinct alopecia scoring systems measure percent hair loss as a common underlying feature, and that an automated algorithm quantifying this percent hair loss from photographs may be designed and applied to all forms of hair loss.