Computer Vision News - November 2022

54 Doctor-to-be Functional Parcellations My thesis also consists of a number of studies that investigate individualized brain mapping using different computational perspectives. In one of the projects, I developed an interactive visual analytics tool to explore and construct individual-specific, functionally-meaningful regions from group level atlases [4 ]. PRAGMA takes an existing reliable prior and allows experts to build functional parcellations that capture individual differences (Fig. 2). It has been shown that the parcellations tailored to individual subjects provide useful information about brain function/dysfunction. Methods to derive these functional parcels commonly rely on fMRI scans that are not routine in clinical practice. In a recent work [5 ], I asked whether tissue properties (cortical thickness, curvature etc.) that are measurable frommore routine clinical scans can tell us about functional regions in the brain. Specifically, I developed a framework to infer individualized functional parcellation from structural MRI measures. The proposed method achieved comparable parcellation accuracy against fMRI derived ground truth labels. In addition, our individual-level parcellations improved homogeneity (similarity of time-courses within parcel) over widely used group parcellations. Figure 2: PRAGMA interface for interactively constructing functional brain parcellations.