Variational regularization of multiple diffusion tensor fields

Abstract

Diffusion Tensor Imaging (DTI) became a popular tool for white matter tract visualization in the brain. It provides quantitative measures of water molecule diffusion anisotropy and the ability to delineate major white matter bundles. The diffusion model of DTI was found to be inappropriate in cases of partial volume effect, such as Multiple Fiber Orientations (MFO) ambiguity. Recently, a variety of image processing methods were proposed to enhance DTI results by reducing noise and correcting artifacts, but most techniques were not designed to resolve MFO ambiguity. In this Chapter we describe variational based DTI processing techniques, and show how such techniques can be adapted to the Multiple Tensor (MT) diffusion model via the Multiple Tensor Variational (MTV) framework. We show how the MTV framework can be used in separating differently oriented white matter fiber bundles.