tracking_tractoracle
Compute the interface between white matter (WM) and grey matter (GM) for seeding and generate a tractogram using TractOracle-IRT, a deep Reinforcement Learning (RL) based tractography algorithm. Through this module, we provide the best(s) pre-trained agent(s) available for tractography using this RL-based method.
Keywords : Diffusion MRI, Tractography, Reinforcement Learning
Inputs
Section titled “Inputs”Input 1
Section titled “Input 1”Format : tuple val(meta), path(wm), path(gm), path(csf), path(fodf)
| Type | Description | Mandatory | Pattern | |
|---|---|---|---|---|
| meta | map | Groovy Map containing sample information e.g. [ id:'test', single_end:false ] | True | |
| wm | file | Nifti white matter probability map. | True | *.{nii,nii.gz} |
| gm | file | Nifti grey matter probability map. | True | *.{nii,nii.gz} |
| csf | file | Nifti cerebrospinal fuild probability map. | True | *.{nii,nii.gz} |
| fodf | file | Nifti image of Spherical harmonic file (fodf). | True | *.{nii,nii.gz} |
Outputs
Section titled “Outputs”Format : tuple val(meta), path(*__tracking.trk)
| Type | Description | Mandatory | Pattern | |
|---|---|---|---|---|
| meta | map | Groovy Map containing sample information e.g. [ id:'test', single_end:false ] | True | |
| *__tracking.trk | file | Tractogram output file. | True | *__tracking.{trk,tck} |
interface_mask
Section titled “interface_mask”Format : tuple val(meta), path(*__interface.nii.gz)
| Type | Description | Mandatory | Pattern | |
|---|---|---|---|---|
| meta | map | Groovy Map containing sample information e.g. [ id:'test', single_end:false ] | True | |
| *__interface.nii.gz | file | Nifti seeding/interface mask for tracking. | True | *__interface.{nii,nii.gz} |
versions
Section titled “versions”Format : path(versions.yml)
| Type | Description | Mandatory | Pattern | |
|---|---|---|---|---|
| versions.yml | file | File containing software versions | True | versions.yml |
Arguments (see process.ext)
Section titled “Arguments (see process.ext)”| Type | Description | Default | Choices | |
|---|---|---|---|---|
| compress | float | The distance threshold for compression of streamlines in mm. | 0.1 | |
| n_actor | integer | Number of actors to use during tracking. This represents the number of streamlines which are propagated in parallel. A higher value will speed up the tracking process, up to a certain point (depending on the number of GPU cores and VRM available). We suggest to use a value between 5000 and 50000. | 10000 | |
| npv | integer | Number of seeds per voxel to use. This directly impacts the total number of streamlines generated. A higher value will result in more streamlines, but also in a longer processing time. | 1 | |
| agent_checkpoint | string | Path to the TractOracle-IRT agent checkpoint that is available within the container. The following checkpoints are available: “sac_irt_inferno”, “sac_irt_hcp”, “crossq_irt_inferno” and “crossq_irt_hcp”. You should not have to change this as the default checkpoint is the best model available for most cases. For more details about those checkpoints, please refer to the TractOracle-IRT documentation: https://github.com/scil-vital/tractoracle_irt. | public://sac_irt_inferno | |
| min_length | float | Minimum length of streamlines in mm. | 20.0 | |
| max_length | float | Maximum length of streamlines in mm. | 250.0 |
| Description | DOI | |
|---|---|---|
| tractoracle-irt | TractOracle-IRT: Exploring the robustness of TractOracle methods in RL-based tractography. | 10.1016/j.media.2025.103743 |
| scilpy | The Sherbrooke Connectivity Imaging Lab (SCIL) Python dMRI processing toolbox. |
Authors
Section titled “Authors”Last updated : 2025-10-20