Deep Learning-Based Vitiligo Activity Evaluation Using Wood's Lamp Imaging: A Clinical Decision Support.

Area of Science:

• Dermatology and Artificial Intelligence
• Medical Imaging Analysis
• Computational Biology

Background:

• Vitiligo presents significant challenges for objective monitoring and treatment evaluation due to its unpredictable nature.
• Current methods for assessing vitiligo progression lack automation and quantitative precision, hindering clinical practice.
• There is a need for advanced tools to support objective follow-up and treatment efficacy assessment in vitiligo management.

Purpose of the Study:

• To design and validate an end-to-end deep learning system for automated vitiligo lesion analysis using Wood's lamp imaging.
• To enable objective localization, longitudinal tracking, and activity assessment of vitiligo lesions.
• To support personalized, stage-aware treatment decision-making through precise disease monitoring.

Main Methods:

• Utilized Mask R-CNN for automated lesion detection and segmentation from Wood's lamp images.
• Employed t-distributed stochastic neighbor embedding (t-SNE) for quantitative pigmentation-state analysis and Vitiligo Disease Activity (VIDA) score for activity evaluation.
• Incorporated risk, correlation, and survival analyses for characterizing disease progression and treatment effects.

Main Results:

• The Mask R-CNN model demonstrated high performance with a Dice coefficient of 90.5% and mean IoU of 83.1% on external validation.
• t-SNE analysis effectively differentiated pigmentation states, correlating with clinical activity.
• Longitudinal analysis revealed treatment benefits for stable lesions, elevated risk for active lesions, and identified age-specific relapse risks.

Conclusions:

• The developed deep learning system provides precise and objective monitoring of vitiligo, outperforming other segmentation models.
• Quantitative analysis of lesion characteristics and disease activity aids in understanding stage-dependent treatment effects.
• This AI-driven approach facilitates personalized treatment strategies and improves clinical decision-making for vitiligo patients.