Edge preserving regularization and tracking for diffusion tensor imaging

Abstract

Two major problems in MR Diffusion Tensor Imaging, regularization and tracking are addressed. Regularization is performed on a variance homogenizing transformation of the tensor field via a nonlinear filter chain to preserve discontinuities. The suitability of the smoothing procedure is validated by Monte Carlo simulations. For tracking, the tensor field is diagonalized and a local bilinear interpolation of the corresponding direction field is performed. The track curves, which are not restricted to the measured grid, are modeled by following stepwise the interpolated directions. The presented methods are illustrated by applications to measured data.