Objective Assessment of Target Vitiligo Areas by Semiautomatic Analysis of Standardized UV Images.

Abstract

Importance: The Target Vitiligo Area Scoring Index (Target-VASI) is the primary tool for evaluating localized vitiligo lesions; however, its uncalibrated estimation in clinical practice is limited by intrarater and interrater variability. While objective methods using tracing papers or imaging systems are available, these approaches are often tedious and rely on images that are not well contrasted.

Objective: To assess the reliability and the performance of a semiautomatic analysis of the lesions performed on standardized UV reflectance pictures.

Design, setting, and participants: This validation study was part of a prospective, double-blind, randomized, placebo-controlled clinical trial conducted between September 2023 and February 2024 at a tertiary care center to evaluate a grafting-based therapy in difficult-to-treat vitiligo lesions. Participants included adult patients with a pair of stable vitiligo target lesions with surface areas between 2 cm2 and 20 cm2. Target-VASI was measured prior to treatment and 12 weeks after treatment by using 3 methods: the image-based system (UV image Target-VASI), tracing paper outlines (tracing paper Target-VASI), and assessments by an experienced clinician (clinician Target-VASI). The image-based system involved semiautomatic analysis of standardized UV reflectance images.

Main outcomes and measures: The Target-VASI estimation methods were compared for correlation, agreement, and repeatability, with relationships scored as weak, moderate, good/strong, and very good/very strong based on results.

Results: A total of 20 adult patients with a pair of stable vitiligo target lesions of surface area between 2 cm2 and 20 cm2 were enrolled. One was lost to follow-up, resulting in 19 evaluable participants included for the analysis. The mean (SD) age was 51.7 (10.7) years, and 12 (63.8%) were female. The mean (SD) vitiligo duration was 21.2 (14.4) years. A total of 61 tracing paper Target-VASIs and 76 UV image and clinician Target-VASIs were available for analysis. UV image Target-VASIs demonstrated strong correlations with tracing paper (r = 0.81 [95% CI, 0.71-0.89]; P < .001) and clinician (r = 0.89 [95% CI, 0.83-0.93]; P < .001) Target-VASIs. The agreement was also good, with mean (SD) biases of 0.81 (0.99) and 0.52 (0.76) for tracing paper and clinician Target-VASIs, respectively. Additionally, an intraclass correlation coefficient of 0.79 (95% CI, 0.60-0.88) indicated moderate to good reproducibility of Target-VASI measurements. Although errors, such as nonmatching lesion regions and oversegmented pigmentation, were identified, the overall UV image Target-VASI accuracy was largely maintained due to the precise detection of depigmented areas.

Conclusions and relevance: In this investigation, the image-based system demonstrated reliable performance, suggesting its possible use in studies or routine clinical practice as a more efficient alternative to conventional approaches.