This research utilizes machine learning (ML) models to predict survival outcomes (12, 24, 36, and 60 months) for patients diagnosed with WHO grade II and III gliomas. Data from the National Cancer Database (NCDB) encompassing 10,001 grade II and 11,456 grade III gliomas were analyzed. LightGBM and Random Forest algorithms yielded the best predictive models, achieving AUROC values ranging from 0.813 to 0.896 for grade II and 0.855 to 0.878 for grade III gliomas. These models are integrated into a user-friendly web application for individualized survival predictions, enhanced by SHAP explanations for improved interpretability. The application aims to integrate predictive analytics into neuro-oncology clinical practice, facilitating personalized prognostication and clinical decision-making.

Mert Karabacak, Pemla Jagtiani, Alejandro Carrasquilla, Isabelle M. Germano, Konstantinos Margetis