An artificial intelligence-based model exploiting H&E images to predict recurrence in negative sentinel lymph-node...

Area of Science:

• Digital pathology
• Melanoma research
• Artificial intelligence in medicine

Background:

• Accurate risk stratification is crucial for negative sentinel lymph node (SLN-) melanoma patients to avoid unnecessary treatments.
• Artificial intelligence (AI) offers potential for improved recurrence risk prediction and adjuvant therapy selection.

Purpose of the Study:

• To develop and validate AI-based models for predicting 2-year recurrence-free status (RFS) in SLN- melanoma patients.
• To assess the performance of AI models utilizing quantitative imaging information from H&E slides.

Main Methods:

• AI models were trained on H&E slides from 71 SLN- melanoma patients, analyzing Regions of Interest (ROIs) with tumor cells alone (TUMOR ROI) and tumor with infiltrating cells (TUMOR+INF ROI).
• Model performance was evaluated using a 5-fold cross-validation and validated on an independent cohort of 23 SLN- melanoma patients.

Main Results:

• AI models analyzing TUMOR ROIs demonstrated higher predictive performance compared to TUMOR+INF ROIs.
• For the investigational cohort, TUMOR ROIs achieved an Area Under the Curve (AUC) of 79.1%, sensitivity of 81.2%, specificity of 70.0%, and accuracy of 73.2%.
• Validation on the external cohort showed AUC of 76.5%, sensitivity of 66.7%, specificity of 70.0%, and accuracy of 70.0% for TUMOR ROIs.

Conclusions:

• This study presents a novel, non-invasive prognostic method for melanoma management.
• The developed AI approach can aid in better defining recurrence risk for SLN- melanoma patients.
• Further development could enhance clinical decision-making regarding adjuvant therapy in melanoma treatment.