AI04: A Novel Machine Learning Software Application for Prognosticating Cutaneous Malignant Melanoma Based on Data from 78,351 Patients from the Surveillance, Epidemiology, and End Results (SEER) Database

The study developed a machine learning (ML) software application to improve prognostication of cutaneous malignant melanoma (CMM) using data from 78,351 patients in the SEER database. The ML model, utilizing Survival eXtreme Gradient Boosting (XGBoost), incorporated 13 clinical and demographic variables, including age, sex, ethnicity, and tumor characteristics, to predict patient survival more accurately than the traditional TNM staging system. The model achieved a concordance index (C-index) of 81% for the training set and 80% for the testing set, outperforming the TNM-based model, which had C-indexes of 66.2% and 66.6%, respectively. The software stratifies patients into five risk categories and provides a Kaplan–Meier plot for survival visualization. The study highlights the importance of integrating additional prognostic factors into clinical tools and suggests further research to validate the application across diverse populations and healthcare systems.