Comprehensive Summary
This article explores the effectiveness of a new AI system in quantifying hair loss severity in patients with alopecia areata using SALT (Severity of Alopecia Tool) scores. Using the already standardized scoring of SALT, the goal is to apply AI to more efficiently and accurately assign scores based on four scalp images (top, back, left, right) of each patient’s head. This study analyzed 188 images from 47 adult patients and divided the hair loss into 4 levels of 0, 0.33, 0.66, and 1. The AI system uses a deep learning model to conduct scalp segmentation and hair loss detection. Each image was scored by both a dermatologist and the AI system. The Intraclass Correlation Coefficient was used to measure the accuracy between the two different raters. The findings revealed that there was high correlation between the dermatologist and AI scores, with the Average ICC of 0.95. The study did mention limitations in using this AI specifically with severe cases or patients with new, white, or thick hair. Overall, this study highlights the large consistency in performance between AI models and dermatologist specialists, specifically emphasizing the potential of using AI models to more efficiently automate SALT scoring.
Outcomes and Implications
With the recent improvements in AI to more efficiently and accurately support doctors, the question of its true validity is critical. In the past, there was evidence of greater reliability on board-certified dermatologists; however, the specific AI models used in the study reflected high accuracy. The study included the importance of continuing to improve the generalizability of this system due to its limitations for specific hair types, colors, and conditions. The study explained that their next plan is to incorporate 3D models of patients’ scalps to more accurately determine SALT scoring and improve the precision and reliability of these SALT scores for improving treatment decisions.