Modified models to distinguish central nervous system demyelinating diseases with brain lesions

Background: Brain lesions in patients with myelin oligodendrocyte glycoprotein antibody-associated disease (MOGAD) are indistinguishable from those with relapsing-remitting multiple sclerosis (RRMS) and aquaporin-4 antibody-positive neuromyelitis optica spectrum disorder (AQP4-Ab NMOSD). Methods: Patients with MOGAD, RRMS, and AQP4-Ab NMOSD with abnormal brain lesions were retrospectively reviewed and divided into training and validation sets. Discriminatory models using brain images and demographics were generated to identify optimal predictors using orthogonal partial least square discriminant analysis after principal component analysis (PCA) of clinico-radiological data without a diagnosis. Constructed models were tested in an independent cohort. Results: PCA of 51 brain scans and demographics from patients (13 MOGAD, 24 RRMS, and 14 AQP4-Ab NMOSD) demonstrated that RRMS was distinct from antibody-mediated conditions. The best predictors between MOGAD and AQP4-Ab NMOSD were poorly demarcated lesions, large abnormalities (both predictive for MOGAD), female sex, disease duration, linear lesions adjacent to the lateral ventricle, and cerebellum involvement (all predictive for MOGAD). Periventricular, ovoid/round, juxtacortical, and callosal lesions; Dawson's fingers; T1 hypointensity (all predictive for RRMS); and fluffy as well as large lesions (for MOGAD) were the best predictors of MOGAD and RRMS. RRMS versus MOGAD and RRMS versus AQP4-Ab NMOSD models exhibited a high predictive value and perfect accuracy (100%), which was validated in an independent cohort. The model of patients with AQP4Ab NMOSD and MOGAD exhibited lower predictive power but still achieved an accuracy of 90%. Conclusions: Brain MRI characteristics combined with demographics enables the distinction of MOGAD from RRMS and AQP4-Ab NMOSD. Fluffy and large lesions are relatively specific MRI characteristics in patients with MOGAD with brain abnormalities in Asian countries.