Project Goal
This project involves using segmentation and classification algorithms to identify major nerve bundles in the brain. We begin with MR data that measures the principal direction of water diffusion in each smallish (2x2x2 mm) voxel. These are called 'Diffusion Tensor Images' (DTI). When water is within a major fiber bundle, it tends to diffuse along the main path of the fiber bundle.
Figure 1: One possible rendering of a slice of a diffusion tensor image of the brain (here each voxel/tensor is mapped to a scalar value representing diffusivity.)
At present, we can visualize the diffusion direction, but there is still a need for algorithms to group the data into the (known) major bundles and to automatically find similar bundles in different people. These are both tasks that we would like to do routinely.
Project Scope
This project involves interfacing with Prof. Brian Wandell of Stanford's Vision Science and Neuroimaging Group and with Dr. Robert Dougherty of the Stanford Institute for Reading and Learning.
Tasks
The project will be accomplished through the following tasks.
- Task 1: Gain access to Diffusion Tensor Image data.
- Task 2: Devise an efficient and robust algorithm for the segmentation of DTI data into known major bundles of fibers.
- Task 3: Given a dataset of DTI images, devise an algorithm to classify similar bundles in different images.
Project Status
Dan Merget and Fiona Wei Ling LokePoint of Contact
Daniel RussakoffMidterm Report
submittedFinal Report
submitted
