Grade and Subtype Classification for Glioma Tumors using Clinical and Molecular Mutations

Gliomas are insidious brain tumors with varying attributes, and their traditional methods of prognosis suffer from notable limitations. In addressing these challenges, we conduct a study on the classification of tumor grades and subtypes, utilizing The Cancer Genome Atlas (TCGA) dataset consisting of 862 data instances. We leverage six feature selection techniques, and several machine learning models, which are scrutinized using metrics like accuracy, F1 score and others. Our experimental results highlight Mutual Information, Chi-Square and Lasso as the best feature selection techniques. We achieve a validation accuracy of 90.70% for grade classification using Mutual Information with an ensemble model of Logistic Regression, KNN and SVM, as well as using two other feature selection-classifier combinations. We further achieve a validation accuracy of 84.88% for subtype classification, using the Bagging classifier with Mutual Information Feature selection. Our framework obtains a higher performance in terms of validation accuracy as compared to the performance of other state of the art classifiers.