ABSTRACT Melasma is a common pigmentary disorder involving melanin deposition and structural alterations. Current diagnostic methods mainly target pigmentation and lack real‐time assessment of histopathology. This study proposes a noninvasive, quantitative evaluation of basement membrane (BM) disruption in melasma using optical coherence tomography (OCT) with deep learning. Cross‐sectional skin images were generated by attenuation coefficient (AC) mapping of OCT B‐scans. An improved Unet (Res‐Att‐Unet), integrating residual and attention modules, was developed for BM segmentation. AC mapping enhanced image contrast, yielding superior BM segmentation over conventional OCT. The proposed model achieved an accuracy of 81.4%, F 1‐score of 83.8%, and IoU of 72.1%. BM loss in melasma (66.0% ± 19.8%) was significantly higher than in perilesional skin (47.2% ± 18.5%, p < 0.001). Longitudinal monitoring revealed a significant BM recovery after tranexamic acid (TXA) treatment. These results indicate that our proposed method can be potentially used in clinic for in vivo BM assessment, aiding in melasma diagnosis.
Liang et al. (2025) studied this question.