Latent dimension linking functional connectivity with post-stroke deficits across multiple domains

Stroke often leads to multiple behavioural impairments. Understanding the neural basis of these deficits is essential for elucidating the mechanisms of functional impairments and optimising therapeutic strategies for stroke patients. Although many studies have revealed that specific behavioural deficits are related to disruptions in distributed functional connectivity across brain networks, these studies typically focus on single behavioural traits, overlooking the multivariate characteristics of deficits after stroke. Recent studies have demonstrated that deficits within and across domains are highly correlated, suggesting a complex many-to-many mapping between brain and behaviour following stroke. Thus, the present study aims to identify meaningful multivariate patterns of functional connectivity-behaviour covariation following stroke. Specifically, we employed a multivariate data-driven approach, partial least squares correlation, to examine the relationships between whole-brain functional connectivity and an extensive array of neurological scores (including motor, attention, verbal memory, spatial memory and language domains) in a large cohort of stroke patients at 2 weeks (n = 81), 3 months (n = 78) and 12 months (n = 74) post-injury. This multivariate analysis revealed a significant latent component (LC) from 2-week post-stroke data, capturing a unique pattern of cognitive deficits across multiple domains. This pattern was strongly associated with widespread network dysfunction, characterized by decreased interhemispheric connectivity and increased intrahemispheric connectivity. Notably, the identified LC was replicated and generalized to stroke data at the 3-month and 12-month time points. Furthermore, we examined whether structural lesion features, including structural disconnection of white matter pathways and grey matter damage, could explain variance in the identified LC. Structural disconnection outperformed grey matter damage, highlighting its critical role in the functional connectivity-behaviour relationship following stroke. Mediation analysis confirmed that structural disconnection serves as the neuroanatomical basis for the association between functional connectivity and deficits. Overall, this study suggests that stroke-induced white matter disconnections are associated with widespread and consistent disruptions in brain network connectivity, which are reflected in a highly correlated behavioural profile. These results provide an integrative insight into the complex relationships among lesions, functional networks, and behavioural outcomes following stroke.