Artificial Intelligence for Lentigo Maligna: Automated Margin Assessment via Sox-10-Based Melanocyte Density Mapping

Area of Science:

• Dermatopathology
• Computational Pathology
• Oncology

Background:

• Lentigo maligna (LM) diagnosis relies on histological evaluation of resection margins, which is challenging and time-consuming.
• Melanocyte density (MD) is a reliable metric for LM assessment.
• Artificial intelligence (AI) offers potential for automating complex diagnostic tasks.

Purpose of the Study:

• To investigate the efficacy of an AI tool in supporting the assessment of LM by measuring melanocyte density (MD).
• To evaluate the AI tool's performance in detecting epidermis and quantifying MD on digitalized slides.
• To determine if the AI tool can enhance the efficiency and reproducibility of LM resection margin assessment.

Main Methods:

• A retrospective single-centre study utilized 86 whole slide images (WSIs) for AI training and 177 WSIs for testing.
• The AI tool was trained to identify the epidermis, measure its length, and calculate MD using Sox-10 stained slides.
• The AI model generated heat maps indicating low, borderline, and high MD, with a defined positive cut-off of ≥30 melanocytes per 0.5 mm.

Main Results:

• The AI model demonstrated high sensitivity (87.84%), specificity (72.82%), and accuracy (79.10%) with an AUC of 0.818 in the test set.
• The AI tool automatically recognized the epidermis and measured MD, comparable to manual counts.
• The model's heat map visualization provided a quick overview of MD distribution across WSIs.

Conclusions:

• The developed AI tool effectively supports dermatopathologists in assessing LM by automating MD measurement.
• This AI-driven approach can significantly improve the efficiency and reproducibility of evaluating resection margins in LM cases.
• The AI model shows promise for integration into routine dermatopathology workflows, aiding in faster and more reliable diagnoses.