Probabilistic diffusion tractography with anatomical priors
Speaker: Anastasia Yendiki , Martinos Center, Harvard Medical School
Date: May 16 2008
Time: 2:00PM to 3:00PM
Location: 32-D507
Host: Polina Golland, CSAIL
Contact: Polina Golland, x38005, polina@csail.mit.edu
Relevant URL: Diffusion tractography uses MR imaging data representing the diffusion of
water molecules through the brain to infer the location and shape of
white-matter fiber bundles. Conventional approaches to tractography use a
deterministic and local model of the diffusion process to step along a
white-matter pathway a few voxels at a time. Such methods are often
confounded by the uncertainty in the diffusion data due to imaging noise and
multiple true diffusion directions. Consequently probabilistic approaches
that attempt to address these issues have been proposed in recent years.
However, most methods often require manual intervention to produce reliable
results, making them less practical for large studies.
In this work we build upon a Bayesian approach to tractography proposed
recently by Jbabdi et al (NeuroImage 2007) that uses a global model of
white-matter pathways. To eliminate reliance on manual intervention and
reduce sensitivity to initialization and end point selection, we
incorporate prior models of the unknown pathways. These models utilize
information on the subject's own anatomy that is routinely available from
a T1-weigthed MRI scan, as well as diffusion and anatomical data from a
set of training subjects. The anatomical information provides constraints
to the possible locations of the tracts with respect to surrounding brain
structures. The resulting method is fully automated and produces tract
solutions that correspond well with expert manual labelings of the same
pathways.
See other events that are part of Biomedical Imaging and Analysis 2007/2008
See other events happening in May 2008
