Show single subject data together with normative values

[valosekj]

This PR adds a generate_figures_with_single_subject.py script to plot single-subject morphometric metrics together with normative values computed from the normative database (spine-generic dataset in PAM50 space).

I tested the script on both females and males.

dcm-zurich/sub-343299 - single compression at C4/C5; male; 0.5×0.5×3.6mm voxel size

Compression:

Commands:

# Label T2w axial spinal cord segmentation using disc labels
sct_label_utils -i sub-343299_acq-axial_T2w_label-SC_mask-manual.nii.gz  -disc sub-343299_acq-axial_T2w_labels-manual.nii.gz -o sub-343299_acq-axial_T2w_label-SC_mask-manual_labeled.nii.gz

# Compute metrics
sct_process_segmentation -i sub-343299_acq-axial_T2w_label-SC_mask-manual.nii.gz -vertfile sub-343299_acq-axial_T2w_label-SC_mask-manual_labeled.nii.gz -perslice 1 -normalize-PAM50 1 -o sub-343299_metrics_PAM50.csv

# Generate figure
python generate_figures_with_single_subject.py -path-HC ${SCT_DIR}/data/PAM50_normalized_metrics -participant-file ${SCT_DIR}/data/PAM50_normalized_metrics/participants.tsv -csv-single-subject sub-343299_metrics_PAM50.csv -single-subject-sex M

dcm-zurich/sub-780875 - single compression at C5/C6; female; 0.5×0.5×3.6mm voxel size

Compression:

Commands:

# Label T2w axial spinal cord segmentation using disc labels
sct_label_utils -i sub-780875_acq-axial_T2w_label-SC_mask-manual.nii.gz  -disc sub-780875_acq-axial_T2w_labels-manual.nii.gz -o sub-780875_acq-axial_T2w_label-SC_mask-manual_labeled.nii.gz

# Compute metrics
sct_process_segmentation -i sub-780875_acq-axial_T2w_label-SC_mask-manual.nii.gz -vertfile sub-780875_acq-axial_T2w_label-SC_mask-manual_labeled.nii.gz -perslice 1 -normalize-PAM50 1 -o sub-780875_metrics_PAM50.csv

# Generate figure
python generate_figures_with_single_subject.py -path-HC ${SCT_DIR}/data/PAM50_normalized_metrics -participant-file ${SCT_DIR}/data/PAM50_normalized_metrics/participants.tsv -csv-single-subject sub-780875_metrics_PAM50.csv -single-subject-sex F

data-multi-subject/sub-tokyoIngenia02 - compression at C6/C7; male; 0.8×0.8×0.8mm voxel size

Compression:

Commands:

# Label T2w spinal cord segmentation using disc labels
sct_label_utils -i sub-tokyoIngenia02_T2w_seg-manual.nii.gz -disc sub-tokyoIngenia02_T2w_labels-disc-manual.nii.gz -o sub-tokyoIngenia02_T2w_seg-manual_labeled.nii.gz

# Compute metrics
sct_process_segmentation -i sub-tokyoIngenia02_T2w_seg-manual.nii.gz -vertfile sub-tokyoIngenia02_T2w_seg-manual_labeled.nii.gz -perslice 1 -normalize-PAM50 1 -o sub-tokyoIngenia02_metrics_PAM50.csv

# Generate figure
python generate_figures_with_single_subject.py -path-HC ${SCT_DIR}/data/PAM50_normalized_metrics -participant-file ${SCT_DIR}/data/PAM50_normalized_metrics/participants.tsv -csv-single-subject sub-tokyoIngenia02_metrics_PAM50.csv -single-subject-sex M

---

TODO:

• also identify other types of compression ("impingement compression" vs "flat compression") and test the script

Figure ref: https://pubmed.ncbi.nlm.nih.gov/27509189/

[jcohenadad]

I really really like this representation of an individual subject on the normative value. Nice work!

[valosekj]

@jcohenadad, what do you think if I implemented the automatic generation of this figure into the sct_process_segmentation function?
If a user calls sct_process_segmentation -normalize-PAM50 1, sct_process_segmentation would output the figure along with a CSV file with metrics in PAM50 (the current output).
This feature could also be advertised in the manuscript.

[jcohenadad]

@jcohenadad, what do you think if I implemented the automatic generation of this figure into the sct_process_segmentation function?

Sure, why not. We need to make sure the fig generation will work on a headless system (ie: using matplotlib OO module) in case people run it on a cluster via SSH. Also, figure naming and output location needs to be figured out. One possibility would be to output that figure via the QC's report. That way, if ppl don't want that figure they simply don't specify -qc, and having it in the QC will be convenient.

[valosekj]

One possibility would be to output that figure via the QC's report. That way, if ppl don't want that figure they simply don't specify -qc, and having it in the QC will be convenient.

Good idea! I discussed the possibility of outputting the PNG figure into the hmtl QC report with @mguaypaq and @joshuacwnewton. The current QC code seems relatively rigid and supports only nii images. Fortunately, refactoring of the QC code currently done by Mathieu (spinalcordtoolbox/spinalcordtoolbox#4204) should offer an easy way of outputting the PNG figure into the hmtl QC report.