Machine Learning and AI Approaches for Classifying Primary Brain Tumours Using Conventional MRI Scans

This study investigates the diagnosis of brain tumours, namely glioblastomas, the most common malignant and incurable brain tumour in adults, by invoking Machine Learning (ML) and AI approaches. We aim to diagnose and infer certain characteristics of brain tumours based on the physical phenomena of the tumour, as gleaned by conventional Magnetic Resonance Imaging (MRI), and by examining inferred identifiers which, which are resident in the cancer's demographic properties. By analyzing the MRI brain scans of living and deceased glioblastoma patients, we determine the maximum dimensions of the tumour in multiple planes, their signal characteristics and their enhancement profiles. In addition, we undertake a texture analysis of the tumours to gather data that is not immediately observable on the MRI scans. The demographic characteristics and genetic biomarkers of the cancer will also be recorded from patients’ medical files. The same information will also be collected for living and deceased patients known to have different types of brain tumours. The information is then analyzed algorithmically to classify patients based on their tumour types using conventional MRI sequences. The future outcome of this research is to determine the survival times of patients with glioblastomas and provide prognostic information in a non-invasive manner, based on the physical characteristics and biological indicators of the tumour.