Table 1.
|
Requirements |
Reproducible |
Processing time |
||||
---|---|---|---|---|---|---|---|
Imaging data | Other | Template registration | DWI preprocessinga | Tractographyb | Total | ||
Whole-brain ACT | T1w DWI |
— | Yes | 39 min | 5 h 45 min | 1 h 12 min | 7 h 36 min |
Manual tractography | T1w DWI |
Neuroanatomical knowledge | No | 39 min | 5 h 45 min | Manual tracking + filtering | 6 h 24 min + manual filtering |
Anatomical pathway model | T1w | — | Yes | 39 min | — | — | 39 min |
Time estimates for a single subject for the three processing pipelines using the minimum data necessary. If a T2w dataset is available then multimodal registration may improve registration accuracy but will increase processing time by ∼20 min. (T1w data for template registration and DWI for streamline generation). To provide more realistic estimates for this table we used a non-HCP unprocessed DWI dataset (2 mm isotropic resolution, 180 directions, multishell [0, 700, 1200, 2800 sec/mm2], reverse-phase-encoded b0 volumes only, acquired on a Siemens 3T Skyra Magnetom). All processing were done using a workstation equipped with a Xeon Gold 5120 CPU (14c/28t @ 2.20GHz).
The DWI preprocessing includes: denoising and Gibbs unringing (MRtrix), as well as correction for EPI distortions, subject motion, and Eddy currents (FSL).
The tractography stage includes generation of functions and for whole-brain ACT only, generation of tissue segmentation map (5TT), and 10 million streamlines (ACT).
ACT, anatomically constrained tractography; DWI, diffusion-weighted imaging; FOD, fiber orientation distribution.