Lin-Ching Chang1, Lindsay Walker2,
Babak Behseta3, Carlo Pierpaoli2
1Department of Electrical
Engineering & Computer Science, the Catholic University of America,
Washington, DC, United States; 2STBB, NICHD, National Institutes
of Health, Bethesda, MD, United States; 3Pediatric &
Developmental Neuroscience Branch, NIMH, National Institutes of Health,
Bethesda, MD, United States
Artifacts are common in diffusion weighted images (DWIs) especially those originating from cardiac pulsation in ungated acquisitions and from subject motion. Neglecting to account for them can result in erroneous diffusion tensor values. The Robust Estimation of Tensors by Outlier Rejection (RESTORE) is an effective method for improving tensor estimation presence of artifactual data points. However, RESTORE relies heavily on data redundancy (large DWIs datasets), which may not be acquired in some clinical settings. This paper proposes a method called informed RESTORE (iRESTORE) that incorporates the notion that physiological noise artifacts are more likely to results in signal drops rather than signal increases to achieve an accurate rejection of artifactual data points in low redundancy DWIs datasets.