High Angular Resolution Diffusion Imaging versus Diffusion Tensor Imaging

One of my main interests in neuroimaging is diffusion MRI fiber tracking. There are multiple methods for tracking including ones based on a diffusion tensor (e.g., FSL’s FDT program). There are also other, more complex mathematical models for describing and analyzing diffusion data that are based on an approach called High Angular Resolution Diffusion Imaging (HARDI) including Q-Ball and mixture of Wishart (MOW). These more complex models allow for the resolution of crossing fibers, a feat traditional tensor models cannot accomplish. Being able to resolve crossing fibers is important because in complex regions of white matter (which likely includes much of the brain), there may be more than one predominant direction fibers are traveling in any given voxel of space.

As an example of the MOW HARDI method of fiber tracking, here are some renderings of tracks starting from various regions of the brain. The MOW tracking was conducted using in-house research software. All other processing and visualization were done using freely available software.

First, here are tracks from a brainstem region of interest in a 64 direction diffusion weighted MRI (2x2x2mm; b=1000). 27 tracking seed points, evenly distributed in a voxel, were calculated for every voxel in the whole brain [typically in our research we seed at least 64 points per voxel]; fiber tracks were then limited to ones passing through the brainstem ROI. In this image, 50% of fibers were skipped for visualization in order to reduce the RAM requirements. Visualization was done with TrackVis.

Here is a rendering of fibers passing through the right hippocampus. There are some noise fibers because of imperfections with the ROI (including image registration problems – the ROI was created using FreeSurfer and then registered to the diffusion image using FNIRT). Again, this is an image created using 27 seed points per voxel; 50% of tracks were visualized.

Below is the same ROI processed using DTI instead of MOW. The MOW method produced 84,000 fibers but the DTI only produced 55,000. The differences are visually obvious. Are they significant? Probably but that’s part of what the research I am doing is focused on.

We can also track between multiple ROIs. As an example of this, here are tracks connecting the right thalamus and the right caudate. Once again, it looks like there are some noise tracks but just a couple.

Now, here is the same brain with the same ROIs (colors of the ROIs are different) but processed using DTI instead of MOW. There is a very noticeable difference between images with the DTI having an aberrant fiber going anteriorly. The DTI method also had only 20000 fibers compared to the MOW’s 23000. That might not seem that significant but this is also a relatively simple tracking with short distance between ROIs and likely not many crossing fibers.

Track count is not the only way we can quantify these tracks. We can create a strength of connection index or a number of other different numbers. With quantification, these images are just pretty pictures and would not be particularly useful for research.

Note: None of the above pictures may be used without express written permission from me.

About Jared Tanner

I have a PhD in Clinical and Health Psychology with an emphasis in neuropsychology at the University of Florida. I previously studied at Brigham Young University. I am currently a Research Assistant Professor at the University of Florida. I spend the bulk of my research time dealing with structural magnetic resonance images of the brain. My specialty is with traditional structural MR images, such as T1-weighted and T2-weighted images, as well as diffusion weighted images. I also look at the cognitive and behavioral functioning of individuals with PD and older adults undergoing orthopedic surgery. Funding for the images came from NINDS K23NS060660 (awarded to Catherine Price, University of Florida). Brain images may not be used without my written permission (grant and software requirements).