Research Projects

This project proposes several deep learning-based novel methods for reducing the overall time complexity and making it feasible from the clinical perspective. Specific objectives: 1. To estimate DWI data of more gradient directions signals 2. To reconstruct multishell diffusion signal data from a single shell data. 3. To map single shell Spherical Harmonics (SH) coefficient data into multi-tissue constrained spherical deconvolution-based Fiber Orientation Distribution Function (fODF). 4. Segmentation would be done at two levels: a coarser (Macro) level where there are two fiber classes, i.e. grey and white matter and a finer (Micro) level where white matter fibers are further classified based on their common pathways (tracts) in connecting brain regions. The overall goal in each step is to reduce time complexity as little as possible, but at the same time, performance should also be achieved state-of-the-art.