AI-assisted quantitative analysis for evaluating melanin distribution in 3D pigmented epidermis-on-a-chip models

Introduction: Abnormal pigmentation plays an important role in various skin diseases and in studies of whitening efficacy.Three-dimensional pigmented epidermis-on-a-chip models provide a crucial in vitro platform for exploring melanin production and regulation in skin. However, dynamic and non-invasive quantitative assessment of melanin distribution remains difficult with traditional histological methods. Methods: In this study, an AI-assisted objective evaluation framework was established for three-dimensional pigmented epidermis-on-a-chip models based on brightfield images. Melanin regions were segmented using the MEM-ViT algorithm, and their morphological features were extracted to build a multi-indicator comprehensive analysis system for determining the "good/poor" status of the model. Results: The results showed 98% consistency between algorithmic predictions and manual annotations, demonstrating the reliability and generalization capability of the proposed method. The framework enabled accurate segmentation of melanin regions and standardized evaluation of model quality without staining. Discussion: This method provides a rapid, non-invasive, and standardized approach for evaluating 3D pigmented epidermis-on-a-chip models. It offers a useful technical pathway for drug efficacy research, whitening mechanism analysis, and objective assessment of skin pigmentation-related disorders.