Abstract Details

Assess feasibility and validity of MRI metrics for MS clinical practice.

Standardized 3D-FLAIR and 3D-T1-weighted MRIs were acquired on 3T scanners (Siemens Healthcare, Erlangen, Germany) as part of routine care at 10 MS PATHS sites, and analyzed using MSPie to estimate brain parenchymal fraction (BPF) and T2 lesion volume (T2LV). MSPie_BPF uses both FLAIR and T1-weighted images. BPF measurement error (RMSE) was measured in patients with ≥4 MRIs. Proportions of patients with BPF RMSE<0.25% were compared across MSPie_BPF, FLAIR_only_BPF, and T1_only_BPF. Spearman correlations were assessed between MSPie metrics and walking speed (WST, 25-foot walk), manual dexterity (MDT, 9-hole peg), and processing speed (PST, symbol digit modalities) tests.

11,385 sets of MSPie metrics were generated for 7,140 unique MS patients with a range of demographic and disease characteristics. For the 317 patients with ≥4 MRIs, MSPie_BPF mean RMSE was 0.17%, and 81% of patients had mean RMSE<0.25%, compared to T1_only_BPF (mean RMSE=0.32% with 43% below cutoff), and FLAIR_only_BPF (mean RMSE=0.22% with 68% below cutoff). Using all available data, MSPie_BPF and T2LV correlated with neuroperformance z-scores (adjusted for age, sex, and education; p<0.0001 for all): WST (r=0.24; r=-0.21, for BPF and T2LV, respectively), MDT (r=0.30; r=-0.29), and PST (r=0.34; r=-0.37).

Standardized MRI acquisitions and tools that can be integrated into radiology workflows, such as MSPie, can be used to generate reliable quantitative metrics for routine care. High reproducibility at the individual patient level supports use of MSPie to monitor MS in clinical practice, while simultaneously creating high-volume MRI data for research. Formal in-clinic testing is ongoing in MS PATHS.