Quantitative MRI phenotypes capture biological heterogeneity in multiple sclerosis patients

Figures

• 

  Figure 1. Timeline of treatment modality, clinical relapses and MRI scans of the longitudinal cohort. Patients included in the longitudinal cohort (N = 33) were scanned at least twice using the same MRI protocol D (72 scans in total: N = 29 had two scans, N = 3 had three scans, and N = 1 had five scans). The radiology reports that were generated in the context of routine clinical follow-up at the University Hospitals Leuven allowed to distinguish between an ‘active’ and ‘stable’ MRI scan at follow-up. Whenever the neuroradiologists mentioned gadolinium enhancement, new lesions or lesions with an increased volume compared to the preceding MRI scan this implied coding of the MRI scan as ‘active’ (Image generated with Rv3.6.1, https://www.R-project.org).

• 

  Figure 2. Longitudinal variation in MRI. Longitudinal evolution of (A–C) median and (D–F) peak height MTR in each tissue class (lesions, NAWM, NAGM) as well as (G) total brain volume, (H) white matter, (J) grey matter and (I) lesion volume. The same color represents each individual patient (N = 33) throughout the graphs. All patients were scanned using the same scanner protocol. MTR magnetization transfer ratio, NAWM normal appearing white matter, NAGM normal appearing grey matter, PH peak height, Tot. total, vol. volume, WM white matter, GM grey matter (Image generated with Rv3.6.1, https://www.R-project.org).

• 

  Figure 3. Contribution of inter- and intra-individual variance in longitudinal MRI data. In the longitudinal cohort of 33 patients we visualized the (A) percentage of variance explained (= adjusted r2) and (B) significance in a generic model (black) including a unique patient identifier code (i.e. inter-individual variation) and time (i.e. intra-individual variation) and a clinical model (grey) including gender, age, disease duration (i.e. known variables of inter-individual variation) and time. The dotted line represents the cut-off for significance of P ≤ 0.05 after correction for multiple testing (P ≤ 0.0025). (C) The relative contributions of inter-individual and intra-individual variation over time to the total variance explained in the generic model. PH peak height, NAWM normal appearing white matter, NAGM normal appearing grey matter, WM white matter, GM grey matter, vol volume (Image generated with Rv3.6.1, https://www.R-project.org).

• 

  Figure 4. Correlation of MTR over tissue classes within patients. Correlation of (A) median and (B) peak height MTR in normal appearing tissues versus median and peak height MTR in lesions. Correlation between (C) median and (D) peak height MTR in NAWM versus NAGM in the cross-sectional study population (N = 213). Adjusted r2 and P-values result from linear regression with covariates age, gender and MRI protocol. The Bonferroni cut-off for significance of P ≤ 0.05 after correction for multiple testing was 0.00047. MTR magnetization transfer ratio, NAWM normal appearing white matter, NAGM normal appearing grey matter, PH peak height (Image generated with Rv3.6.1, https://www.R-project.org).

• 

  Figure 5. Association of MS genetic risk score with MRI traits. Correlation of (A–C) median and (D–F) peak height MTR across tissues as well as (G–J) volumetric measurements with MS genetic risk score calculated based on known HLA (N = 22) and non-HLA (N = 133) risk variants in the cross-sectional study population (N = 209). P-values were calculated from linear regression including age, gender and protocol as covariates. NAGM normal appearing grey matter, NAWM normal appearing white matter, PH peak height, GM grey matter, WM white matter, vol. volume, MS multiple sclerosis, HLA human leukocyte antigen (Image generated with Rv3.6.1, https://www.R-project.org).