Cluster-based parallel processing of DT-MRI acquisitions for real time probabilistic fiber tracking
Using acquisition data from Diffusion Tensor Magnetic Resonance Imaging (DT-MRI), an MRI variant that can map the diffusion of water, medical researchers have been able to image neural bundles in the brain, known as white matter tracts. Fiber tractography has been used extensively for clinical diagnosis of many neurological disorders and is used by neurosurgeons for surgical planning. Dipy, an open source project, provides a rich suite of fiber tracking algorithms accessible to any medical center in the world. Computing the fiber tracks using its probabilistic models, however, are computationally expensive. My project this summer is to integrate standard parallelization frameworks into dipys probabilistic models. I plan on leveraging industry standard cloud computing tools to accelerate the computation of these fiber tracts. The goals of my project are to enhance dipys probabilistic models for richer image processing, to provide real-time fiber tracking analytics to UCSF surgical planners, and to provide a clean interface for DT-MRI researchers to benchmark and compare their tractography analysis against dipys probabilistic models.
Message to Sponsor
- Major: Computer Science, Bioengineering Minor
- Sponsor: Rose Hills
- Mentor: Youngho Seo, Radiology and Biomedical Imaging (UCSF)