Skip to content

Measure of the averaged cross-sectional area (cord CSA) between C2 and C3 of the spinal cord with UK Biobank Brain MRI dataset.

License

Notifications You must be signed in to change notification settings

sct-pipeline/ukbiobank-spinalcord-csa

Folders and files

NameName
Last commit message
Last commit date

Latest commit

Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 

Repository files navigation

Cord CSA on UK biobank brain MRI database

This repository contains the analysis pipeline to normalize spinal cord cross-sectional area (CSA).

πŸ‘‰ Please cite this work if you use it or if you are inspired by it:

BΓ©dard S, Cohen-Adad J. Automatic measure and normalization of spinal cord cross-sectional area using the pontomedullary junction. Frontiers in Neuroimaging [Internet]. 2022;1. Available from: https://www.frontiersin.org/articles/10.3389/fnimg.2022.1031253

Spinal cord CSA was computed on T1w MRI scans for 804 participants from the UK Biobank database. In addition to computing cross-sectional at the C2-C3 vertebral disc, it was also measured at 64 mm caudal from the pontomedullary junction (PMJ). The effect of various biological, demographic and anatomical factors was explored by computing Pearson’s correlation coefficients. A stepwise linear regression found significant predictors; the coefficients of the best fit model were used to normalize CSA.

Measure of the averaged cross-sectional area (CSA) between C2 and C3 of the spinal cord with UK Biobank Brain MRI dataset.

Table of contents


Data collection and organization

Uk Biobank database

The brain MRI data of UK biobank follows the DICOM convention. The spinal cord of the processed brain MRI images is cut off. Because the purpose of this project is to measure the CSA between C2 and C3 of the spinal cord, the raw MRI images for T1w structural images and for T2w FLAIR are used as an input of the pipeline:

  • T1_orig_defaced.nii.gz
  • T2_FLAIR_orig_defaced.nii.gz

The analysis pipeline was applied on T1w images only.

The raw images have gradient distortion, correction will be applied in the preprocessing steps of the analysis pipeline.

The DICOM dataset is under: duke:mri/uk_biobank

Data conversion: DICOM to BIDS

For this project, a BIDS standard dataset is used. A conversion of DICOM to BIDS is necessary for the UK Biobank dataset. The data from the DICOM dataset in the BIDS standard for this project have the following correspondance for each subjects:

  • T1_orig_defaced.nii.gz in the BIDS standard is sub-XXXXXXX_T1w.nii.gz
  • T2_FLAIR_orig_defaced.nii.gz in the BIDS standard issub-XXXXXXX_T2w.nii.gz

To convert the DICOM dataset in a BIDS structure for this project, run the following line:

curate_project.py -path-in <path_DICOM_dataset> -path-output <path_BIDS_dataset>

The BIDS datset with raw data is under: data.neuro.polymtl.ca:datasets/uk-biobank. The dataset resulting from preprocessing is under data.neuro.polymtl.ca:datasets/uk-biobank-processed.

Here is an example of the BIDS data structure of uk-biobank-processed:

uk-biobank-processed
β”‚
β”œβ”€β”€ dataset_description.json
β”œβ”€β”€ participants.json
β”œβ”€β”€ participants.tsv
β”œβ”€β”€ README
β”œβ”€β”€ sub-1000032
β”œβ”€β”€ sub-1000083
β”œβ”€β”€ sub-1000252
β”œβ”€β”€ sub-1000498
β”œβ”€β”€ sub-1000537
β”œβ”€β”€ sub-1000710
β”‚   β”‚
β”‚   └── anat
β”‚       β”œβ”€β”€ sub-1000710_T1w.json
β”‚       β”œβ”€β”€ sub-1000710_T1w.nii.gz
β”‚       β”œβ”€β”€ sub-1000710_T2w.json
β”‚       └── sub-1000710_T2w.nii.gz
└── derivatives
    β”‚
    └── labels
        └── sub-1000710
            β”‚
            └── anat
                β”œβ”€β”€ sub-1000710_T1w_seg-manual.nii.gz  <---------- manually-corrected spinal cord segmentation
                β”œβ”€β”€ sub-1000710_T1w_seg-manual.json  <------------ information about origin of segmentation
                β”œβ”€β”€ sub-1000710_T1w_labels-manual.nii.gz  <------- manual vertebral labels
                β”œβ”€β”€ sub-1000710_T1w_labels-manual.json
                β”œβ”€β”€ sub-1000710_T1w_pmj-manual.nii.gz  <------- manual pmj label
                β”œβ”€β”€ sub-1000710_T1w_pmj-manual.json
                β”œβ”€β”€ sub-1000710_T2w_seg-manual.nii.gz  <---------- manually-corrected spinal cord segmentation
                └── sub-1000710_T2w_seg-manual.json
 

Acquisition parameters

Scanner: Siemens Skyra 3T running VD13A SP4 with a standard Siemens 32-channel RF receive head coil

T1-weighted structural imaging

Resolution: 1x1x1 mm
Field-of-view: 208x256x256 matrix

T2-weighted FLAIR structural imaging

Resolution: 1.05x1x1 mm
Field-of-view: 192x256x256 matrix

Analysis pipeline

This repository includes a collection of scripts to analyse a BIDS-structured MRI dataset.

Dependencies

MANDATORY:

  • SCT 5.3.0 for processing
  • gradunwarp for gradient correction
  • ANTs for processing
  • Python 3.7 for statistical analysis

OPTIONAL:

  • ITK-SNAP for correcting cord segmentations

    NOTE: Make sure to add ITK-SNAP to the system path:

    • For Windows, select the option during installation.
    • For macOS, after installation, go to Help->Install Command-Line Tools.

Installation

Download this repository:

git clone https://github.com/sct-pipeline/ukbiobank-spinalcord-csa.git

Install:

cd ukbiobank-spinalcord-csa
pip install -e ./

Note on gradient distortion correction

A coeff.grad associated with the MRI scanner used for the data is necessary if it has not been applied yet. In this project, the gradient distortion correction is done in preprocess_data.sh with gradunwarp and Siemens coeff.grad file.


Preprocessing

Preprocessing generates a dataset with gradient distortion correction.

First, initialize shell variable with the path to the folder with the coeff.grad file:

PATH_GRADCORR_FILE=<path-gradcorr>

Launch preprocessing:

sct_run_batch -jobs -1 -path-data <PATH-DATA> -path-output ~/ukbiobank_preprocess -script preprocess_data.sh -script-args $PATH_GRADCORR_FILE

The results to use as the new dataset wil be in ~/ukbiobank_preprocess/data_processed/.

Processing

Processing will generate spinal cord segmentation, vertebral labels, pmj label and compute cord CSA. Specify the path of preprocessed dataset with the flag path-data.

Launch processing:

sct_run_batch -jobs -1 -path-data <PATH_DATA> -path-output ~/ukbiobank_results/ -script process_data.sh

Or you can launch processing with a config file instead by using the flag -config and by adjusting the file config_sct_run_batch.yml according to your setups. See sct_run_batch -h to look at the available options. To launch processing:

sct_run_batch -config config_sct_run_batch.yml

Quality control

After running the analysis, check your Quality Control (qc) report by opening the file ~/ukbiobank_results/qc/index.html. Use the "search" feature of the QC report to quickly jump to segmentations or labeling issues.

1. Assess quality of segmentation, vertebral and pontomedullary junction (PMJ) labeling

If segmentation or labeling issues are noticed while checking the quality report, proceed to manual correction using the procedure below:

  1. In QC report, search for "deepseg" to only display results of spinal cord segmentation, search for "vertebrae" to only display vertebral labeling and "pmj" to only display PMJ labeling.
  2. Review segmentation and spinal cord and PMJ labeling, note that the segmentation et vertebral labeling need to be accurate only between C2-C3, for cord CSA.
  3. Click on the F key to indicate if the segmentation/label is OK βœ…, needs manual correction ❌ or if the data is not usable ⚠️ (artifact). Two .yml lists, one for manual corrections and one for the unusable data, will automatically be generated.
  4. Download the lists by clicking on Download QC Fails and on Download Qc Artifacts.

Note: Proceed to QC separately for cord segmentation and vertebral labeling to generate 2 separate lists.

The lists will have the following format:

.yml list for correcting cord segmentation:

FILES_SEG:
- sub-1000032_T1w.nii.gz
- sub-1000083_T2w.nii.gz

.yml list for correcting vertebral labeling:

FILES_LABEL:
- sub-1000032_T1w.nii.gz
- sub-1000710_T1w.nii.gz

.yml list for correcting pontomedullary junction (PMJ) labeling:

FILES_PMJ:
 - sub-1000032_T1w.nii.gz
 - sub-1000710_T1w.nii.gz
  • FILES_SEG: Images associated with spinal cord segmentation
  • FILES_LABEL Images associated with vertebral labeling
  • FILES_PMJ Images associated with PMJ labeling

For the next steps, the script uk_manual_correction loops through all the files listed in .yml file and opens an interactive window to either manually correct segmentation, vertebral or PMJ labeling. Each manually-corrected label is saved under derivatives/labels/ folder at the root of PATH_DATA according to the BIDS convention. Each manually-corrected file has the suffix -manual. The procedure is described bellow for cord segmentation and for vertebral labeling.

2. Correct segmentations

For manual segmentation, you will need ITK-SNAP and this repository only. See installation instructions and dependencies.

Here is a tutorial for manually correcting segmentations. Note that the new QC report format with interactive features (βœ…/❌/⚠️) is not included in the tutorial.

IMAGE ALT TEXT

Run the following line and specify the .yml list for cord segmentation with the flag -config:

uk_manual_correction -config <.yml file> -path-in ~/ukbiobank_results/data_processed -path-out <PATH_DATA>

After all corrections are done, you can generate a QC report by adding the flag -qc-only- to the command above. Note that SCT is required for generating QC report.

3. Vertebral labeling

Note that manual labeling uses SCT and the QC report is generated automatically.

Here is a tutorial for manual vertebral labeling:

IMAGE ALT TEXT

Run the following line and specify the .yml list for vertebral labeling with the flag -config:

uk_manual_correction -config <.yml file> -path-in ~/ukbiobank_results/data_processed -path-out <PATH_DATA>

To create disc labels, click at the posterior tip of the disc for C1-C2, C2-C3 and C3-C4 as shown in the following image:

readme_labels

4. PMJ labeling

Note that manual PMJ labeling uses SCT and the QC report is generated automatically.

Run the following line and specify the .yml list for PMJ labeling with the flag -config:

uk_manual_correction -config <.yml file> -path-in ~/ukbiobank_results/data_processed -path-out <PATH_DATA>

To create PMJ label, click at the posterior tip of the pontomedullary junction (PMJ) as shown in the following image:

image

See here for more examples.

Upload the manually-corrected files

A QC report of the manually corrected files is created in a zip file. To update the dataset, add all manually-corrected files derivatives/labels/, and include the qc zip file in the body of the PR. See our internal procedure for more details.

Add automatic segmentations to derivatives/ folder

After all segmentations are manually QC-ed, you can add them to the derivatives/ by running again the script manual_correction.py and adding the flag -add-seg-only:

uk_manual_correction -config <.yml file> -path-in ~/ukbiobank_results/data_processed -path-out <PATH_DATA> -add-seg-only

The automatic segmentations that did not require manual correction (files in .yml file) are added to the derivatives/ folder. You can upload the folder following the instructions specified in Upload the manually-corrected files.

Re-run the analysis

After all the necessary segmentation and labels are corrected, re-run the analysis (sct_run_batch command in Processing). If manually-corrected files exist, they will be used instead of proceeding to automatic segmentation and labeling. Make sure to put the output results in another folder (flag -path-output) if you don't want the previous results to be overwritten.


Statistical analysis

Generate datafile:

To generate a data file with the CSA results from process_data.sh and a .csv file including the fields from the UK Biobank dataset, run the follwing line:

uk_get_subject_info -path-data <PATH_DATA> -path-output ~/ukbiobank_results/ -datafile <FILENAME>

Compute statistical analysis

To compute the statistical analysis of cord CSA results, use uk_compute_stats. If the datafile ouput of uk_get_subject_info is not data_ukbiobank.csv, add the flag -dataFile <FILENAME>. Run this script in /results folder or specify path to folder that contains output files of analysis pipeline in with -path-output flag. The flag -exclude points to a yml file containing the subjects to be excluded from the statistical analysis:

uk_compute_stats -path-output ~/ukbiobank_results/ -dataFile <DATAFILE> -exclude <EXCLUDE.yml>

The output of uk_compute_statshas the following data structure:

ukbiobank_results
β”‚
β”œβ”€β”€ data_processed
β”œβ”€β”€ log
β”œβ”€β”€ qc
└── results
    β”œβ”€β”€ csa-SC_c2c3.csv
    β”œβ”€β”€ csa-SC_pmj.csv
    β”œβ”€β”€ data_ukbiobank.csv
    └── stats_results
        β”œβ”€β”€ metrics
        |   β”œβ”€β”€ comparasion_c2c3_pmj
        |   |   └── scatterplots_c2c3_pmj_csa.png
        |   β”œβ”€β”€ scatter_plots
        |   |   β”œβ”€β”€ Age.jpeg
        |   |   β”œβ”€β”€ Brain GM volume.png
        |   |   β”œβ”€β”€ Brain WM volume.png
        |   |   ...
        |   β”œβ”€β”€ corr_table.csv
        |   β”œβ”€β”€ corr_table_and_pvalue.csv
        |   β”œβ”€β”€ corr_table_pvalue.csv        
        |   β”œβ”€β”€ stats_csa.csv
        |   β”œβ”€β”€ scatter_plot_age_brain_vol.png
        |   β”œβ”€β”€ scatter_plot_age_thalamus_vol.png
        |   └── stats_param.csv
        └── models
            β”œβ”€β”€ age
            |   β”œβ”€β”€ coeff
            |   |  β”œβ”€β”€ coeff_linear_fit.csv
            |   |  └── coeff_quadratic_fit.csv
            |   β”œβ”€β”€ summary
            |   |   β”œβ”€β”€ summary_linear_fit.txt
            |   |   └── summary_quadratic_fit.txt
            |   └── quadratic_fit.png
            β”œβ”€β”€ model_1
            |   └── CSA_PMJ
            |      β”œβ”€β”€ coeff
            |      |   β”œβ”€β”€ coeff_fullLin_CSA_PMJ.csv
            |      |   └── coeff_stepwise_CSA_PMJ.csv
            |      β”œβ”€β”€ residuals
            |      |    β”œβ”€β”€ res_plots_fullLin_CSA_PMJ.png
            |      |    └── res_plots_stepwise_CSA_PMJ.png
            |      β”œβ”€β”€ summary
            |      |   β”œβ”€β”€ summary_fullLin_CSA_PMJ.txt
            |      |   └── summary_stepwise_CSA_PMJ.txt
            |      └── compared_models.csv
            β”œβ”€β”€ model_2
            |  └── ...
            β”œβ”€β”€ sex
            |   └── violin_plot.csv
            └──norm_COV.csv
            

The coefficient files (coeff_*.csv) are included in SCT to normalize CSA.

About

Measure of the averaged cross-sectional area (cord CSA) between C2 and C3 of the spinal cord with UK Biobank Brain MRI dataset.

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Contributors 4

  •  
  •  
  •  
  •