My PhD dissertation work focused on the development of image analysis techniques applied to MRI for quantifying early neurodegeneration in Alzheimer’s disease (AD). AD diagnosis is significantly challenged by the fact that multiple pathologic processes frequently co-occur in the brains of patients with AD. My work aims to disentangle this pathological heterogeneity by combining information from high resolution ex vivo MRI data, and serial digital histopathology to isolate changes in the brain associated specifically with AD. Such analyses will be used to develop MRI biomarkers that will empower clinical trials in early AD to detect therapeutic effects earlier and with greater precision than current methods allow.

More specifically, my project focuses on the construction of a computational atlas of the medial temporal lobe, one of the earliest brain regions affected by AD, using machine learning and computer vision techniques applied 3D ex vivo MRI and histology data. The template provides a common reference space across specimens which can be used to characterize anatomical variability and effects of AD-related pathology on MTL structure. Conventional approches for atlas construction fail to handle ultra-high resolution MRI and geometric variability in the medial temporal lobe. Therefore, to address these challenges, I develop and validate customized image segmentation and registration pipelines using classical image analysis and deep learning techniques and concepts such as shape awareness and geometric modelling.