This work aims to develop and evaluate new tools for non-invasive analysis of white matter fiber bundles in the brain. Diffusion weighted magnetic resonance imaging (MRI) provides information on fiber orientation because the diffusion of tissue water is faster parallel than perpendicular to axons. In addition, the diffusion measurement is sensitive to cellular properties such as cell type, size, and volume fraction. Diffusion tensor imaging (DTI) has been widely used to assess white matter development and pathology, as well as to reconstruct fiber paths in the brain. However, it has become clear that the technique has two major limitations. First, image noise produces both noise and bias in the estimated diffusion tensor. This complicates comparisons of DTI parameters between subjects, and leads to errors in estimated fiber paths. Second, conventional DTI is based on the assumption that a single tensor describes the diffusion in each image voxel. The single-tensor model is inappropriate in regions of the brain with complicated fiber structure (e.g., crossing or splaying fibers). An alternative method, 'q-space' imaging, can be used to quantify diffusion using very few assumptions, but requires impractically long scans for most clinical or pediatric studies. Extended q-space experiments show that the single-tensor model is generally inappropriate, but at present there are no established alternatives suitable for routine clinical use. The studies in this proposal aim to evaluate and mitigate the effects of noise and partial volume averaging in DTI. Specifically, we propose the following: (1) to test theoretical predictions of the effects of noise on the diffusion tensor and estimated fiber paths, and to quantify the performance of a tensor denoising algorithm. (2) To develop high spatial resolution DTI using navigated, multiple shot echo planar imaging. (3) To develop a time-efficient approach to DTI that uses a multiple-tensor model of diffusion in each voxel, and test its reliability in phantom and human studies. (4) To develop approaches to fiber tracking and segmentation using the new multiple-tensor information, and test these in phantom and human subjects. We expect that the resulting methods will significantly improve the sensitivity of DTI studies. Overall this work will help to define the capabilities and limitations of fiber characterization using MRI, which represents a powerful new approach for studying the structure of neural tissue.
Showing the most recent 10 out of 13 publications
