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

Area of Science:

• Biotechnology
• Dermatology
• Artificial Intelligence

Background:

• Abnormal skin pigmentation is key in skin diseases and whitening efficacy studies.
• 3D pigmented epidermis-on-a-chip models are vital for studying melanin production in vitro.
• Traditional histological methods struggle with dynamic, non-invasive melanin assessment.

Purpose of the Study:

• To develop an AI-assisted framework for objective evaluation of 3D pigmented epidermis-on-a-chip models.
• To enable non-invasive, quantitative assessment of melanin distribution.
• To standardize the quality assessment of skin models.

Main Methods:

• Established an AI-assisted evaluation framework using brightfield images.
• Employed the MEM-ViT algorithm for melanin region segmentation.
• Developed a multi-indicator system for model quality analysis.

Main Results:

• Achieved 98% consistency between AI predictions and manual annotations.
• Demonstrated the reliability and generalization of the AI method.
• Enabled accurate melanin segmentation and standardized model evaluation without staining.

Conclusions:

• The AI framework offers a rapid, non-invasive, and standardized method for evaluating 3D pigmented skin models.
• This approach supports drug efficacy research, whitening mechanism studies, and skin disorder assessment.
• Provides a valuable technical pathway for objective skin pigmentation analysis.