Inter-rater reliability of atherosclerotic plaque enhancement on 7 T vessel wall MRI: Comparison with quantitative analysis

Résumé fourni par la source

Objective Intracranial plaque enhancement (IPE) is a potential biomarker of plaque vulnerability but lacks a standardized definition. While subjective assessment may be prone to observer variability, a voxel-based quantification method can detect subtle signal intensity (SI) changes. This study aimed to compare the inter-rater reliability of subjective IPE evaluation with that of a voxel-based quantification method. Methods Patients with stroke due to intracranial atherosclerosis were prospectively imaged using 7 T vessel wall magnetic resonance imaging. Two adjudicators independently assessed IPE, with discrepancies resolved by a third, experienced neuroradiologist. IPE distribution was quantified using a semiautomated method based on SI measurements from post-contrast T1-weighted images. The mean enhancement ratio was calculated for each plaque to assess gadolinium enhancement. Inter-rater reproducibility and the concordance between subjective and objective assessments were evaluated. Regression analysis was performed to identify plaque morphological features that influenced the reliability of subjective IPE adjudication compared to the objective method. Results We analyzed 75 plaques from 41 patients. Inter-rater agreement for IPE adjudication was poor ( κ = 0.34). Following consensus, 43% of plaques (32/75) were classified as enhancing, whereas 57% (43/75) were non-enhancing. The agreement between subjective and objective IPE assessments was moderate ( κ = 0.40, p < 0.001). Subjective assessment was more likely to reliably adjudicate IPE for plaques exhibiting >50% stenosis (odds ratio (OR), 1.04; p = 0.02) and a positive remodeling index (OR, 2.20; p = 0.02). In contrast, it was less reliable when evaluating posterior circulation plaques (OR, 0.23; p = 0.01). Conclusion Subjective assessment of IPE demonstrated poor inter-rater agreement and only moderate concordance with voxel-based quantification following consensus.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.