Survival Prediction of Glioma Patients Using Machine Learning

Gliomas are one of the most common types of brain tumors with varying prognoses, treatment responses, and clinical outcomes. Predicting the survival of patients with glioma is a challenging task due to the complexity and variety of parameters that contribute to the outcome, such as genetic, molecular, and clinical factors. Machine learning has been increasingly used in medical applications in recent years due to its ability to under-stand or approximate complex relationships of multidimensional data. In this work, we aim to predict the survival of glioma patients by training and comparing the performance of different machine-learning approaches on multi-feature patient data. Our tests included simple models such as support vector machines SVMs and decision trees, as well as multiple ensemble models. Using a small private dataset of 81 samples, our best-performing models: Random Forest, and Stacking classifier achieve an accuracy of 0.97, fl-score of 0.97, precision of 0.97, and recall of 0.97. Finally, we utilize explainability methods to further describe the effects of different features of patient data on the predicted outcome.