TORTOISEV4: Reimagining the NIH diffusion MRI processing pipeline.

Diffusion MRI (dMRI) data suffer from a number of artifacts, including, but not limited to, low SNR, Gibbs ringing, bulk subject motion, within volume motion, eddy-current distortions, susceptibility-induced EPI distortions, and ghost artifacts. Appropriate pre-processing of diffusion-weighted images prior to model fitting is vital for accurate quantitative analysis. Over the years, the nature of dMRI data has evolved (smaller voxel sizes, significantly larger number of volumes and b-values, wider variety of acquisition paradigms, etc.) as have the required processing tools. Additionally, very large multi-site dMRI studies, on potentially uncooperative subjects (young children, geriatric populations, patients with movement disorders, etc.), have increased the necessity for dMRI processing pipelines that are fast, robustly capable of handling a variety of artifacts/distortions, and that have summary reporting capabilities to pinpoint problematic data. TORTOISE (Tolerably Obsessive Registration and Tensor Optimization Indolent Software Ensemble) (www.tortoisedti.org) has been redesigned, made adaptable and significantly enriched to meet these needs.