MS100Public

MATLAB analysis code for:

Tract-based Quantitative MRI for Resolving the Clinico-Radiological Paradox in Multiple Sclerosis

Abdullah O*, Wen P*, Abdelrazeq AW*, Matrosova M*, Brylev L, Ahmad A, Bryukhov V, Melcher D, Rokers B.

Overview

This repository contains the public analysis scripts used to generate the tract-based quantitative MRI figures and tables for the manuscript. The code operates on a local BIDS-formatted multiple sclerosis dataset plus derivative spreadsheets and MRI maps. Patient-level clinical and imaging data are not included in this repository.

Requirements

• MATLAB R2024b or later
• Statistics and Machine Learning Toolbox
• Image Processing Toolbox, for NIfTI I/O
• Parallel Computing Toolbox, for scripts that use parpool and parfor

Setup

1. Clone this repository.

2. Edit configs/config.json and set bidsDir to the root of your local BIDS dataset.

3. From MATLAB, add the repository to the path:

   addpath(genpath('/path/to/MS100Public'));

4. Create the local output directory used by the regression scripts:

   mkdir(fullfile('/path/to/MS100Public', 'outputs'));

Repository Structure

MS100Public/
  configs/
    config.json              Data paths and output filenames
    tract_names.txt          White matter tract names
    tract_groups/            Tract-to-group assignments
  helpers/                   Shared loading, plotting, and statistics helpers
  outputs/                   Local generated outputs; not tracked
  *.m                        Analysis and figure scripts

Main Figures

Figure	Script	Description
1	fig1_clinical_distributions.m	Clinical score distributions and radar plots
2	Not code-generated	MRI acquisition and preprocessing schematic
3	Rendered externally	Classical versus tract-based lesion topography
4	revisionExtras/code/fig4_plain_matched89.m; original radar scripts: fig4_lesionload_tract_radar.m, fig4_lesionload_classical_radar.m	Reproducible 89-person bar and six-radar MATLAB figures
5	fig5_T1_radar.m, fig5_FA_radar.m, fig5_MD_radar.m, fig5_MTR_radar.m	Lesion and normal-appearing white matter qMRI radar plots
6	rev_fig6_r2_comparison.m	R-squared comparison for tract-based versus classical models

Tables

Table	Script	Description
1	table1_verify.m	Demographic and clinical summary statistics
2	paper_multi_bi_ridge_all.m, rev_qMRI.m	Binary ridge regression summaries for EDSS and MSPro

Supplementary Materials

Item	Script	Description
Figure S1	figS1_correlation_heatmaps.m	Correlation heatmaps with Bonferroni correction
Table S3	rev_qMRI.m	Ridge regression with demographic covariates
Table S4	paper_bh_p_correction.m, rev_p_correct.m	Bonferroni-Holm corrected p-values
Table S5	rev_lesionload.m	Lesion-load ridge regression

Configuration Notes

• configs/tract_names.txt lists the tract names used by the analysis.
• configs/tract_groups/*.txt defines the tract groupings used in group-level analyses.
• configs/config.json controls the BIDS root, derivatives folders, output spreadsheet names, and plotting colors.
• Generated regression outputs are written to outputs/ inside this repository.
• Generated qMRI spreadsheet outputs are written to <bidsDir>/<statsDir>/ unless a script is run in test mode.

Figure 4 plain plots on the matched imaging cohort

revisionExtras/code/fig4_plain_matched89.m reads GroupTractLesionLoad.xlsx and icometrixLesionLoad.xlsx from the configured stats directory. It restricts the classical file to the 89 participants in the tract file (RRMS 49, PPMS 17, SPMS 23; sub-RR047 is the sole classical-only row). It produces two separate MATLAB figures: a Figure 4A bar chart and a Figure 4B six-polar-panel plot. The manuscript's Illustrator layout is a separate, local styling step.

From MATLAB, add revisionExtras/code to the path and choose a new output directory for the two vector PDFs:

addpath('/path/to/MS100Public/revisionExtras/code');
fig4_plain_matched89('/path/to/new/figure4-plain-output');

The function refuses an existing output directory. If the configured volume is not mounted, pass the tract and classical workbook paths as its second and third arguments. Figure 4A shows mean ± sample SD in mL, with a 0–50 mL axis; its tract bars sum the four tract-region lesion volumes and are not a unique-voxel union. Figure 4B retains real zeros and shows regional mean ± SEM. Lesion-volume fields are converted from mm³ to mL, and each radial maximum is 1.1 times its largest plotted mean + SEM, matching the original MATLAB radar calculation.

Data Availability

The clinical and imaging data cannot be shared publicly due to patient privacy. De-identified data may be available from the corresponding author on reasonable request, subject to ethics committee approval.

License

This code is provided for academic research purposes. Please cite the paper if you use this code.