Research Report on Changes in Functional Connectivity in Post-Stroke Epilepsy (PSE) Patients Based on Source-Level EEG and Graph Theory

Research Background

Epilepsy has various etiologies, including idiopathic, congenital, head trauma, central nervous system infections, brain tumors, neurodegenerative diseases, and cerebrovascular diseases. Among these, cerebrovascular diseases account for about 11% of all epilepsy cases and are the most common cause of epilepsy in the elderly. Furthermore, post-stroke epilepsy (PSE) is one of the common complications in stroke patients, with 3% to 30% of stroke patients potentially developing PSE. Risk factors for stroke-related epilepsy include cortical involvement, hemorrhagic transformation, early seizures, younger age at onset, the severity of the stroke (e.g., high NIHSS score), and alcohol abuse.

Network science and graph theory are considered to have significant potential in understanding brain functions. Graph metrics can reflect the integration and segregation properties of networks, thus being widely used in various neurological disease studies, including epilepsy. However, although existing studies have shown significant changes in brain network topology in epilepsy patients compared to healthy controls, no research has focused on functional connectivity changes in PSE patients using source-level EEG analysis.

Research Origin

This paper was written by Dong Ah Lee, Taeik Jang, Jaeho Kang, Seongho Park, and Kang Min Park, all from the Department of Neurology at Busan Paik Hospital, Inje University College of Medicine, South Korea. The article has been published in the journal “Brain Topography” and was officially accepted on March 28, 2024.

Research Methods

Research Subjects

This study prospectively recruited 30 PSE stroke patients and 35 non-PSE stroke patients with approval from the hospital’s ethics committee. These patients had no history of epilepsy or seizures before the stroke and had no psychiatric disorders, developmental disorders, or other serious medical histories. We collected the clinical characteristics of the patients, such as age, gender, cause of the stroke (based on TOAST classification), location and side of the stroke, NIHSS score, presence of hemorrhagic transformation, comorbidities (e.g., hypertension, diabetes, dyslipidemia, atrial fibrillation, etc.), time interval from stroke onset to EEG examination, and the presence of bilateral tonic-clonic seizures.

EEG Data Collection and Preprocessing

EEG data were collected while the patients were in a resting state, with eyes closed but awake. The same type of EEG equipment was used, and 23 electrodes were placed according to the international 10-20 system. The sampling frequency was 250 Hz, and each recording lasted for at least 30 minutes. Data were manually reviewed, and artifact-free 5-second periods with alpha activity and no epileptiform discharges were preprocessed.

Source Estimation and Functional Connectivity Matrix Creation

Source estimation was conducted using the Brainstorm program with minimum norm imaging, and nodes were defined based on the Desikan-Killiany Atlas (68 regions of interest in the cortex). Phase-locking value (PLV) was used to measure brain synchronization. Data were processed according to EEG frequency bands (Alpha, Beta, Theta, Delta, low Gamma, and high Gamma), and a 68×68 weighted connection matrix was established for each group.

Graph Theory Analysis

Graph theory analysis was conducted using the Braph program, obtaining functional connectivity measures from the connection matrix, such as average strength, radius, diameter, average eccentricity, characteristic path length, average clustering coefficient, transitivity, and small-world index. These connectivity measures were compared between PSE and non-PSE patients using non-parametric permutation tests.

Statistical Analysis

Clinical characteristics were compared between the two groups using chi-square tests, independent sample t-tests, or Mann-Whitney tests. Functional connectivity measures were statistically analyzed using non-parametric tests with 1,000 permutations, with significant results adjusted using ∆ notation for p-values.

Research Results

Comparison of Clinical Characteristics

Results showed no significant differences in age, gender, cause of stroke, side and location of stroke, NIHSS score, or comorbidity characteristics between the two groups. However, the EEG detection time interval was significantly longer in the PSE group than in the non-PSE group (29.0 weeks vs. 2.0 weeks, p<0.001), indicating that PSE patients were examined for a longer period post-stroke.

EEG Spectral Power Analysis

No significant differences were found in relative spectral power between the two groups in the Delta, Theta, Alpha, Beta, low Gamma, and high Gamma frequency bands.

Comparison of Functional Connectivity Measures

Graph theory analysis revealed that the radius and diameter significantly increased in PSE patients in the Beta and high Gamma bands, and the radius increased significantly in the low Gamma band. In the high Gamma band, PSE patients showed a significant decrease in average strength, average clustering coefficient, and transitivity. This indicates a decrease in both segregation and integration of brain networks in PSE patients, especially noticeable in the Beta and Gamma bands.

Conclusion and Significance of the Study

This study is the first to use source-level EEG analysis and graph theory to investigate changes in functional connectivity in PSE patients. The results indicate significant changes in brain network segregation and integration in PSE patients compared to non-PSE stroke patients, mainly in high-frequency (Beta and Gamma) bands. These findings suggest that the strengthened epilepsy mechanisms in PSE are closely related to changes in brain functional connectivity, providing important references for understanding the pathogenesis of PSE and potential clinical intervention strategies.

Study Highlights

The highlights of this study include providing new insights into brain functional connectivity in PSE patients through source-level EEG analysis and using graph theory to comprehensively quantify brain network characteristics across different frequency bands. Even with a small sample size and a single-center study, the research clearly elucidates the mechanisms of PSE epilepsy and proposes potential clinical application values, laying the foundation for future large-scale, multi-center longitudinal studies.

Study Limitations

Due to the relatively small sample size and single-center nature of this study, these results should be further validated in larger-scale, multi-center studies. Additionally, differences in EEG testing intervals may influence the results, and future study designs need to control and consider this factor.

Summary

The study on changes in functional connectivity in post-stroke epilepsy patients based on source-level EEG and graph theory reveals significant changes in brain network topology in PSE patients. This provides valuable insights for understanding the mechanisms of epilepsy onset in PSE and potential clinical intervention strategies.