Functional brain connectivity in young adults with post-stroke epilepsy

Abstract

Approximately 1 in 10 young stroke patients (18-50 years) will develop post-stroke epilepsy, which is associated with cognitive impairment. While previous studies have shown altered brain connectivity in patients with epilepsy, little is however known about the changes in functional brain connectivity in young stroke patients with post-stroke epilepsy and their relationship with cognitive impairment. Therefore, we aimed to investigate whether young ischaemic stroke patients have altered functional networks and whether this alteration is related to cognitive impairment. We included 164 participants with a first-ever cerebral infarction at young age (18-50 years), along with 77 age- and sex-matched controls, from the Follow-Up of Transient Ischemic Attack and Stroke patients and Unelucidated Risk Factor Evaluation study. All participants underwent neuropsychological testing and resting-state functional MRI to generate functional connectivity networks. At follow-up (10.5 years after the index event), 23 participants developed post-stroke epilepsy. Graph theoretical analysis revealed functional network reorganization in participants with post-stroke epilepsy, in whom a weaker (i.e. network strength), less-integrated (i.e. global efficiency) and less-segregated (i.e. clustering coefficient and local efficiency) functional network was observed compared with the participants without post-stroke epilepsy group and the controls (P < 0.05). Regional analysis showed a trend towards decreased clustering coefficient, local efficiency and nodal efficiency in contralesional brain regions, including the caudal anterior cingulate cortex, posterior cingulate cortex, precuneus, superior frontal gyrus and insula in participants with post-stroke epilepsy compared with those without post-stroke epilepsy. Furthermore, participants with post-stroke epilepsy more often had impairment in the processing speed domain than the group without post-stroke epilepsy, in whom the network properties of the precuneus were positively associated with processing speed performance. Our findings suggest that post-stroke epilepsy is associated with functional reorganization of the brain network after stroke that is characterized by a weaker, less-integrated and less-segregated brain network in young ischaemic stroke patients compared with patients without post-stroke epilepsy. The contralesional brain regions, which are mostly considered as hub regions, might be particularly involved in the altered functional network and may contribute to cognitive impairment in post-stroke epilepsy patients. Overall, our findings provide additional evidence for a potential role of disrupted functional network as underlying pathophysiological mechanism for cognitive impairment in patients with post-stroke epilepsy.

Keywords: network analysis; post-stroke epilepsy; resting-state functional MRI; young stroke.

Graphical Abstract

Figure 1

Flow chart of the study population. fMRI, functional magnetic resonance imaging; ICH, intracerebral haemorrhage; IS, ischaemic stroke; TIA, transient ischaemic attack.

Figure 2

The pipeline for processing resting-state fMRI (rs-fMRI). (A) T₁-weighted images and rs-fMRI were obtained from participants and controls. (B) After fMRI image pre-processing, 82 brain regions (41 in each hemisphere) were defined. (C) For each brain region, an average BOLD time course was extracted. (D) These BOLD time courses were used for the construction of connectivity matrices for each participants, and (E) brain network measures were calculated, on both global and regional levels. BOLD, blood oxygen level dependent; fMRI, functional magnetic resonance imaging.

Figure 3

Whole-brain network analysis. Box plots of the functional network properties, divided by group (n = 241). Corresponding P-values computed with ANCOVA testing, adjusted for age and sex, are added below.

Figure 4

Regional network analysis. Brain regions (in red) of the contralateral (e.g. ‘healthy’) hemisphere, showing lower clustering coefficient (A), local efficiency (B) and nodal efficiency (C) in participants with PSE compared with participants without PSE (P-uncorrected < 0.05), for which general linear models were used (n = 158). After FDR correction for multiple comparisons, none of the brain regions remained significant.