Spatially regularized low-rank tensor approximation for accurate and fast tractography

Johannes Gruen, Samuel Groeschel, Thomas Schultz

Abstract: Highlights•We introduce two novel spatially regularized tractography algorithms based on low-rank tensor approximation.•Integrating low-rank approximation into an unscented Kalman filter (UKF) increases accuracy and, at the same time, significantly reduces computational effort.•Joint tensor approximation accounts for neighboring voxels that contain the same tract and achieves even more complete reconstructions from single seed regions than the UKF.•Benefits over the state of the art are demonstrated on data from the Human Connectome Project, the 2015 ISMRM tractography challenge, and a clinical tumor dataset.