Advancing Glioblastoma Treatment Through AI-Driven Radiomics: A Comparative Study of Feature Selection and Machine Learning Techniques

Tumour characterisation in the field of glioblastoma radiomics presents a substantial obstacle. Glioblastoma, renowned for its elevated severity and frequent incidence, need advanced diagnostic techniques for effective treatment. Radiomics is a crucial approach in this field that utilises quantitative analysis of medical pictures. This work conducts a comprehensive comparative assessment of several techniques for selecting relevant features and machine learning classifiers specifically designed for glioblastoma radiomics. The text provides a comprehensive assessment and comparison of the effectiveness of several feature selection methods, including Random Forest feature importance, Correlation-Based Feature Selection, Mutual Information, L1 Regularisation (Lasso), and XGBoost feature significance. The research concurrently examines the performance of many classifiers, such as K-Nearest Neighbour, Quadratic Discriminant Analysis, and Multi-layer Perceptron, among others. The main objective is to identify the most powerful combinations that improve the capacity to forecast accurately and reliably, therefore making a significant contribution to the advancement of personalised medical therapies for brain tumours.