In this cross-sectional study we assess inter-rater level of agreement (LoA) on PRL identification using a T2*-weighted (T2*-w) 3-dimensional echo-planar imaging (3D-EPI)-derived phase images at 3 Tesla (3T) and a single-echo (SE) gradient-echo (GRE)-derived phase and susceptibility-weighted imaging (SWI) at 7 Tesla (7T).
Paramagnetic rim lesions (PRLs) are an important magnetic resonance imaging (MRI) biomarker in multiple sclerosis (MS), with proposed inclusion in updated diagnostic criteria. Consequently, an accurate identification of PRLs using clinically accessible, vendor-provided MRI sequences is vital for routine practice.
Twenty-eight newly diagnosed, treatment-naïve people with MS, clinically isolated syndrome (CIS), or radiologically isolated syndrome (RIS) underwent brain MRI at both 3T and 7T. The 3T protocol included T2-weighted fluid attenuated inversion recovery (T2w-FLAIR) sequence and T2*w-3D-EPI GRE sequences. The 7T protocol included T2w-FLAIR and SE-GRE sequences. PRLs were prospectively assessed by three independent raters using the 3T T2*w-3D-EPI derived phase map, and 7T SE-GRE derived phase and SWI maps, following the North American Imaging of MS criteria. Inter-rater and rater-to-consensus LoA were evaluated using Cohen’s kappa (κ). Sensitivity and specificity of each sequence were calculated against 7T SWI.
PRLs were more frequently detected on 7T SWI (n=37, 5.2%) and 7T phase (n=35, 5.0%) than on 3T phase (n=25, 3.5%) (p<0.001 for both comparisons). Inter-rater LoA was substantial on 3T phase (κ=0.66–0.74), substantial to almost perfect on 7T phase (κ=0.76–0.82) and almost perfect on 7T SWI (κ=0.82–0.90). Compared to 7T SWI, 3T phase showed substantial LoA (κ=0.75), with 99.6% specificity and 66.7% sensitivity. The LoA between 7T phase and SWI was nearly perfect (κ=0.96).
3T phase imaging demonstrates substantial agreement and high specificity relative to 7T SWI, supporting its clinical utility as a practical alternative for PRL detection.