Directional correlation in white matter tracks of the human brain

Purpose: To describe a technique for the detection of distinct brain fibers in sets of magnetic resonance (MR) diffusion tensor imaging (DTI) data.

Materials and methods: MR-DTI can be used for a tractography of brain fibers presuming a data set of high spatial resolution and high signal to noise. A less demanding technique for the visualization of discrete brain fiber bundles involves segmentation. By using a region-growing algorithm, those voxels that have a direction similar to that of the major eigenvector in neighboring voxels of a data set can be marked. It has been shown recently by Mori et al (1) that this technique can be successfully applied to data from a single slice of a mouse brain. In this study, the segmentation technique was applied with modifications to multislice DTI data from the human brain.

Results: A distinct segmentation of various brain fiber bundles could be achieved by the use of a two-step algorithm. In the first step, voxels within large fiber tracts-such as corticofugal tracts (e.g., corticospinal tract) and the optic radiation-were segmented by starting the region-growing algorithm in the corpus callosum (CC) and erasing this major structure from the data set. In the second step, remaining voxels were segmented by the same algorithm; this revealed a good assignment of the similarly oriented fibers derived by segmentation to the anatomically given brain lobes. This two-step procedure was successfully applied to DTI data of six healthy volunteers.

Conclusion: The segmentation technique for DTI data proposed by Mori et al (1) for data from mouse brains can be applied to multislice data from the human brain by using a two-step algorithm including a masking of the major fiber tracts.