Mathematical Modelling of Malignant Transformation in Low Grade Gliomas and survival prediction with XGBoost.

Motivation: MRI features associated with malignant transformation (MT) have low specificity and circumscribed by diagnostic errors. Goal(s): The objective of this study was to develop a non-linear machine learning model to predict overall survival (OS) quantified by known time-to-death data. Approach: We aimed to assess the prognostic significance of the model inputs and identify the most critical determinants of survival through the model training loop. Results: The RMSE and MAE values in the testing set indicate that the model predicts time to death with an average deviation of ~2.27 years. Impact: Prediction of malignant transformation of low grade gliomas will allow precision treatment decisions and personalized medicine.