Jun Zhang, Qian Jiang, Qiang Chen, Bin Hu, Liuqing Chen

Department of Dermatology, Wuhan No. 1 hospital, Wuhan, Hubei, People’s Republic of China

Correspondence: Bin Hu, Email binhu88@126.com Liuqing Chen, Email chlq35@126.com

Background: Melasma is a prevalent pigmentary disorder characterized by treatment resistance and high recurrence. Existing assessment methods like the Melasma Area and Severity Index (MASI) are subjective and prone to inter-observer variability.
Objective: This study aimed to develop an AI-assisted, real-time melasma severity classification framework based on deep learning and clinical facial images.
Methods: A total of 1368 anonymized facial images were collected from clinically diagnosed melasma patients. After image preprocessing and MASI-based labeling, six CNN architectures were trained and evaluated using PyTorch. Model performance was assessed through accuracy, precision, recall, F1-score, AUC, and interpretability via Layer-wise Relevance Propagation (LRP).
Results: GoogLeNet achieved the best performance, with an accuracy of 0.755 and an F1-score of 0.756. AUC values across severity levels reached 0.93 (mild), 0.86 (moderate), and 0.94 (severe). LRP analysis confirmed GoogLeNet’s superior feature attribution.
Conclusion: This study presents a robust, interpretable deep learning model for melasma severity classification, offering enhanced diagnostic consistency. Future work will integrate multimodal data for more comprehensive assessment.

Keywords: melasma, deep learning, convolutional neural networks, MASI, clinical decision support