FibGrid: Characterizing Anatomical Pathways in Hypergrid for Direct Tract Segmentation

Diffusion MRI tractography is employed to delineate the white matter tracts in the brain, which plays a pivotal role in investigating brain development and psychiatric disorders. The segmentation of fibers into coherent bundles holds paramount importance for brain function investigation and preoperative strategizing in medicine. We propose a novel registration-free method for segmenting whole-brain fibers, called as FibGrid. This approach encodes fibers into a combination of four independent features, resulting in a concise representation of the fiber structures. Subsequently, segmentation is performed within low-dimensional feature spaces defined by these encoded features, followed by classification using a fiber classifier. Compared to the two traditional methods, FibGrid exhibits stable and promising identification of anatomical pathways. Generalization experiments across the dataset and tract atlas, our method is significantly more efficient than others, consistently delivering good segmentation outcomes. The code is available at https://github.com/x-y-u/FibGrid.