Neuroanatomical subtypes of Long COVID reveal distinct cognitive profiles alongside longitudinal brain changes

BackgroundLong COVID is characterized by persistent symptoms following SARS-CoV-2 infection, including cognitive deficits among its most disabling manifestations. However, their neurobiological basis is poorly understood. Previous neuroimaging studies reported structural brain alterations, but few have integrated cognitive assessment, network-based analyses, and longitudinal imaging to identify neurobiological subtypes. MethodsWe studied 42 individuals with persistent post-COVID cognitive symptoms and 14 matched healthy controls using cognitive assessment and structural MRI. Multilayer brain networks were constructed from regional morphometric measures, and hierarchical clustering identified patient subgroups. Cross-sectional and longitudinal comparisons of brain structure, cognition, and function were performed, with exploratory brain-behavior correlations over six months. ResultsClustering identified two subgroups with distinct structural patterns. Compared with controls, one subgroup showed reduced gray matter density in cerebellar lobules VIIIa/VIIIb and the putamen (p < 0.05, TFCE-corrected), whereas the other showed no significant alterations. Cognitive differences between clusters did not survive multiple-comparison correction; however, several measures showed medium-to-large effect sizes (d = 0.70-1.01), suggesting meaningful cognitive differences requiring confirmation in larger cohorts. Longitudinal analyses revealed increased medial frontal gray matter density (p < 0.05) associated with visuospatial/executive performance. ConclusionsThese exploratory findings suggest that post-COVID cognitive symptoms are associated with heterogeneous neurobiological profiles rather than a uniform pattern of impairment. Structural alterations involving cerebellar, striatal, and frontal regions may reflect distinct neuroanatomical phenotypes. Longitudinal findings suggest medial frontal structural reorganization with functional relevance. These findings support data-driven stratification for characterizing neurobiological heterogeneity in Long COVID and provide a foundation for future hypothesis-driven studies.