Artificial Intelligence Enhances Facial Vitiligo Assessment With High Accuracy

A new artificial intelligence (AI) algorithm has demonstrated high accuracy in assessing facial vitiligo, offering a faster and more objective method for evaluating disease severity and treatment response, according to a study published in the Journal of Investigative Dermatology.

Vitiligo is the most common acquired skin depigmenting disorder and significantly impacts patients' quality of life. While clinical tools such as the Vitiligo Area Scoring Index (VASI) and the Vitiligo Extent Score are widely used, they can be time-consuming. Recent advances in machine learning have enabled AI-driven image analysis for skin conditions; however, research on neural network-based assessment of vitiligo has been limited.

“The objective of this study was to build and evaluate a neural network allowing the detection of vitiligo lesions on photographs with a special interest in lesions located on the face,” explained Dirk Hillmer, BRIC (BoRdeaux Institute of onCology), INSERM UMR1312, Team 5, University of Bordeaux in Bordeaux, France, and coauthors.

The AI model processed clinical images by extracting facial landmarks and creating a vitiligo mask to quantify depigmented areas. In a comparative analysis of 69 images from 59 patients, the AI’s assessment closely matched that of 3 expert physicians, with interrater reliability scores of 0.80 (95% CI, 0.71–0.87) and intrarater reliability of 0.81 (95% CI, 0.78–0.90). AI assessments correlated well with individual raters (0.67–0.70) and effectively tracked changes in vitiligo lesions over time. However, minor discrepancies occurred, particularly in cases of small or extensive lesions, likely due to image quality and human interpretation variability.

“In conclusion, this validated AI model to assess facial vitiligo could become a useful tool in clinical practice and clinical research trials,” concluded the study authors.

Reference

Hillmer D, Merhi R, Boniface K, et al. Evaluation of facial vitiligo severity with a mixed clinical and artificial intelligence approach. J Invest Dermatol. 2024;144(2):351-357.e4. doi:10.1016/j.jid.2023.07.014