Visualizing the brain structure with a DT-MRI minimum spanning tree

Visualizing the human brain using diffusion tensor magnetic resonance imaging (DT-MRI) data has been a key technique to study the structure of the human brain and its connectivity. The challenge is to find a method that best exploits the data and serves as a model for visualization and connectivity analysis. This paper presents a novel method of visualizing the human brain structure with a minimum spanning tree using DT-MRI data. The human brain is modeled as a graph in which each vertex represents a brain voxel and each edge represents connectivity between a pair of neighboring brain voxels, resulting in each vertex having 26 weighted connections with adjacent voxels. The weight of an edge is calculated from the DT-MRI data with a higher weight assigned to an edge that are more likely aligned with nerve fiber trajectories. The method then grows a minimum spanning tree representing paths of the nerve fiber bundles. The resultant minimum spanning tree is consistent with the known anatomical appearances of the human brain. As the minimum spanning tree representing the human brain is a global deterministic model with well-defined connectivity between voxels in the brain, it can serve not only as a deterministic visualization of the human brain but also as an instrument for connectivity analysis. In addition, this method overcomes several problems present in previous methods such as tracking termination in traditional fiber tracking and meaningless streamlines in stochastic connectivity mapping.