Meeting Banner
Abstract #1904

Filtered Multi-Tensor Tractography Using Free Water Estimation

Christian Baumgartner1, Ofer Pasternak2, Sylvain Bouix2, Carl-Fredrik Westin3, Yogesh Rathi2

1Information Technology and Electrical Engineering, ETH Zrich, Zrich, Switzerland; 2Psychiatry Neuroimaging Laboratory, Harvard Medical School, Boston, MA, United States; 3Laboratory of Mathematics in Imaging, Harvard Medical School, Boston, MA, United States

In this work, we describe a tractography method that simultaneously estimates multiple diffusion tensors and an isotropic component, we term as free water. The model consists of three Gaussian tensors, one of which represents isotropic diffusion of free water, that are fitted to the DWI-signal using an unscented Kalman filtering framework. By this means, each estimation is guided by those previous, resulting in an inherent regularization of the tracts. The proposed method can be useful in tracing fibers through edema or lesions, where traditional tractography algorithms fail.

Keywords

able account acquisition addition affected anatomical apparent appears assume basis become becomes blue body bounded brain broken complete component computing confidence constant constrain constraints constructed contrasts contribution correctness cortical covariance degenerate deterministic diagonal diffusion drastically dynamics edema eigenvalues electrical elements ellipsoidal employ enables enforce engineering equation error estimation examine example expectations extending fail fiber fibers filter filtered finally find forming forward fraction framework free full gradient green guided healthy hence human identity inherent intervention isotropic iteration laboratory lastly lateral least lesions linear lobe local loop match mathematics matrices matrix medical mixture model models moreover neural newly next notice obeys original originating orthonormal parietal partly pick positive principal problem produces programming propagate proposed psychiatry quadratic reconstruction remaining represent represents requirements rich sample sampling school shades shape since solver space squared squares step still subject surrounded takes tensor tensors term traced tracing tract trans transition transversal tumor unconstrained unscented useful water wish written