Generative Artificial Intelligence Tools: Evaluating Ways to Automate Your SALT Scoring of Alopecia Areata

    Radhika Gupta, Griffin Stockton Hogrogian, Hojjat Salmasian, Sarah Mehlmann, Leslie Castelo-Soccio, George Cotsarelis, Michelle Oboite, Albert C. Yan
    TLDR Generative AI tools like GPT-4o can effectively automate SALT scoring for alopecia areata, matching clinician accuracy.
    The study evaluated the ability of GPT-4o, a generative AI tool, to automate the generation of SALT scores for alopecia areata (AA) using 104 de-identified 4-view images of patients' scalps. The AI-generated scores were compared to clinician-derived scores using intraclass correlation coefficients (ICC) and concordance correlation coefficients (CCC). The results showed high concordance between GPT-4o and both in-person and image-based provider assessments, with ICC and CCC values ranging from 0.815 to 0.950. These findings suggest that generative AI tools like GPT-4o could serve as effective adjuncts in SALT scoring, reducing subjectivity without requiring additional training.
    Discuss this study in the Community →