Visualizing the Intrinsic Geometry of Dynamic Human Brain Connectomes

Understanding how brain regions are interconnected is an important topic within the domain of neuroimaging. The progresses in techniques not invasive such as fMRI (functional Magnetic Resonance Imaging) and DTI (Diffusion Tensor Imaging), allow now to quickly collect maps of brain structure and function with an high level of details. These data contribute to create what is usually referred to as a connectome, that is, the comprehensive map of neural connections. Given the context, brain connectomics have become visible as field which aims is to understand these complete maps of brain connectivity using complex computational models. As it happens, the availability of connectome data allows for more interesting questions to be asked and more complex analyses to be conducted. In this thesis work I present BRAINtrinsic, a new web- based analytics tool that allows user to interactively explore in 3D the intrinsic geometry of the connectome. The brain’s intrinsic geometry is the result of brain data that has been transformed through a dimensionality reduction step, such as multidimensional scaling (MDS), isomap, or t-distributed stochastic neighbor embedding (t-SNE) techniques. BRAINtrinsic is fully compatible with the Oculus Rift device since it has been designed from the beginning to be completely compliant with the virtual reality world. The BRAINtrinsic visualization tool has been evaluated through a real set of case studies, demonstrating its effectiveness in aiding domain experts for a range of neuroimaging tasks. Particularly, a visualization tool for these datasets would help neuroradiologists to have a deeper understanding of the meaning of graph-based metrics when applied to the connectome network as well as to provide more accurate diagnosis in the clinical cohorts of psychiatric and neurological disease such as bipolar depression and Alzheimer’s disease.