Association of Cortical Atrophy Patterns With Clinical Phenotypes and Histopathological Findings in Patients With Rasmussen Syndrome

Automated MRI analyses have identified variable patterns of cortical atrophy in Rasmussen syndrome. In this study, we aim to identify imaging phenotypes of Rasmussen syndrome, to clinically characterize these phenotypes, and to validate this imaging-based approach through histopathologic analysis. For this retrospective case-control study, individuals with Rasmussen syndrome diagnosed according to the European Consensus Statement and at least one 3D T1-weighted MRI scan (<20 years after onset) were identified from the University Hospital Bonn (1995-2023). Healthy controls were selected from databases at the University Hospital Bonn, Charité University Hospital Berlin, and the Human Connectome Project. Disease epicenters, describing brain regions highly connected to atrophy regions, were mapped individually using network-based atrophy modeling. Subtypes were identified through k-means clustering. Neuropsychological test results and results from neuropathologic analyses of biopsies were ascertained, and correlations between subtype-specific atrophy maps and normative maps (enhancing neuro imaging genetics through meta analysis [ENIGMA] and neuromaps toolbox) were used to characterize atrophy profiles and epicenter susceptibility. The study incorporated 54 individuals with Rasmussen syndrome (median age at MRI: 18 years, range 2-61, 65% female) and 270 healthy individuals (median age at MRI: 26.5 years, range 3-61, 49% female). Four distinct atrophy subtypes were identified (temporoparietal, centrotemporal, frontal, and bilateral). Individuals with the centrotemporal subtype were younger at onset (median 5.5 years) than individuals with temporoparietal (median 11.5 years, p = 0.02) and frontal (median 6 years, p = 0.02) subtypes. Most severe neuropsychological impairment was observed for the temporoparietal and frontal subtypes. In the temporoparietal and frontal subtypes, atrophy occurred preferentially in hubs (r = -0.28, p = 0.006; r = -0.30, p = 0.02). Disease epicenter susceptibility was associated with higher cortical thickness (r = -0.57, p = 0.005), lower myelin content (r = 0.47, p = 0.02), lower cerebral blood flow (r = 0.42, p = 0.03), lower blood volume (r = 0.57, p = 0.006), and lower oxygen metabolism (r = 0.47, p = 0.01). Brain biopsies showing strong inflammation were taken from likely epicenters, whereas biopsies with weaker inflammation came from less likely epicenters (p = 0.04). Using Rasmussen syndrome as a model, we validate imaging-based mapping of individual disease epicenters with histopathologic evidence. With further validation, network-based mapping of individual disease epicenters could potentially be used in Rasmussen syndrome to guide biopsy site selection, inform treatment decisions, and improve outcome prognoses.