Projects:DiffusionTensorImageFiltering

Back to NA-MIC_Collaborations, Utah Algorithms

Diffusion Tenson Image Filtering

Coronal slice from a noisy diffusion tensor image (left). The same slice after applying our DTI filtering method (right).

We are developing new denoising methods for diffusion tensor MRI. These methods are based on physical noise models in DT-MRI.

Description

We have implemented several filtering methods for DT-MRI, including our new method and also several methods from the literature. One goal is to determine whether it is best to filter the estimated tensor fields or the original diffusion weighted images. The method that we developed filters the original diffusion weighted images and takes into account the physical properties of the imaging noise. We are comparing this method with others in the literature, including methods that filter the estimated tensor fields. Our preliminary findings are that it is advantageous to filter the DWIs and to include a physical model of the noise.

Key Investigators

• Utah: Saurav Basu, Tom Fletcher, Ross Whitaker
• BWH: Sylvain Bouix, Doug Markant, Adam Cohen, Marc Niethammer, Marek Kubicki, Mark Dreusicke, Martha Shenton

Links

Project Week Results: Jan 2006