Checks and stats
Once the preprocessing and the tracking are done, scilpy also contains scripts to separate your data into known bundles (Recobundles), vizualize your results (ex, scil_visualize_bundles_mosaic.py), compute statistics on your bundles and so on.
QA/ QC (Quality assurance, Quality check) – tractoflow
We have a QC pipeline which creates a summary of the quality of your tractoflow results.
How to install it?
Your can clone and install the dmriqcpy repository:
git clone https://github.com/scilus/dmriqcpy.git cd dmriqcpy pip install -e .If you want to use it as a nextflow (see Nextflow: users for more information.), you can use the singularities available with the releases on Github (click on “Assets” under the chosen release). To download the nextflow:
git clone https://github.com/scilus/dmriqc_flow.git
How to run it?
Here is an example of how to run the nexflow version. If you need to use it on a cluster (ex, Beluga), you can check our Tractoflow tutorial.
# -profile: one of input_qc, tractoflow_qc_light or tractoflow_qc_all. nextflow run PATH_TO_DMRIQC_FLOW/main.nf --input TRACTOFLOW_RESULTS -resume \ -profile tractoflow_qc_all
How to use it?
This produces a QC report for each important section of the tractoflow. For instance, you may check the quality of the registration step, or compare the FA of all your subjects. You can check each html file by double-clicking on it in your file explorer. Check the interrogation mark on the top right corner for a list of shortcuts. The QC typically raises warning when a subject is very different from the others. You can then check each tab, each subject for a visual inspection. Scroll through the subjects with the arrows. For each subject, you can then decide if the quality level was acceptable (hit 1,2 or 3 for pass, warning or fail), and leave comments in the comment box.
We will soon add here a list of characteristic you should expect to see for each metric, to help beginners develop an expert eye when looking at processed data.
QA/ QC (Quality assurance, Quality check) – recobundlesX
To assess the quality of your bundles, the best way is to check the data visually using the mosaic, as described on the RecobundlesX page. However if you possess the ground truth for your data, you may use tractometry for your statistics.
Tractometry
If you have a ground truth, you can compute statistics for each bundle using the metrics that were introduced in the tractometer. You can find the definition of each metric (ex., VB, IB, VC, IC, OL, OR) on the tractometer website. To compute them, you can use scilpy’s script scil_evaluate_bundles_pairwise_agreement_measures.py.