Title of Thesis: Classification-based Glioma Diffusion Modeling Classification-based Glioma Diffusion Modeling

Gliomas are diffuse, invasive brain tumours that originate from a single glial cell and infiltrate through adjacent healthy tissue as the number of tumour cells exponentially increases. The goal of this thesis is to study glioma diffusion and to propose a classification model that would shed some light on glioma growth patterns. We introduce a 3D classification-based diffusion model, CDM, that predicts how a brain tumour will grow at a voxel-level on the basis of features specific to the patient and the tumour, and attributes of that voxel and its neighbours. We use Machine Learning algorithms to learn this probabilistic general model, based on the observed growth patterns of gliomas from other patients. We demonstrate that our learned CDM model can, in many cases, predict glioma growth more effectively than two standard models: uniform radial growth across all tissue types and another that assumes faster diffusion in white matter. We study CDM numerically and analytically on clinical data.