PhD Defense: Image Analysis and Deep Learning Techniques for the Detection and Characterisation of Unruptured Intracranial Aneurysms

Unruptured Intracranial Aneurysms (UIAs) are a bulging in the wall of a blood vessel, which if ruptured will lead to bleeding in the brain with potentially devastating consequences. To prevent rupture, robust and reliable growth and rupture risk assessment of UIAs should be made from medical brain scans to aid in clinical treatment decision making.

In this thesis, image analysis techniques for the detection and characterisation of UIAs from time-of-flight magnetic resonance images (TOF-MRAs) are presented. Automatic UIA detection and segmentation techniques are presented and objectively evaluated in an international biomedical image analysis challenge. Based on volumetric segmentation, automatic characterisation of the volume and morphology of UIAs was investigated, to aid in aneurysm growth classification. Combining techniques and using vessel surface meshes, a model was developed for indicating UIAs at high risk of future growth. The methods presented are a major step forward in computer-assisted growth assessment of UIAs, to help UIA treatment decision making and hopefully contribute to patient outcome.