Asymmetric distribution of enlarged perivascular spaces in centrum semiovale may be associated with epilepsy after acute ischemic stroke

Abstract Objective To investigate the factors influencing enlarged perivascular space (EPVS) characteristics at the onset of acute ischemic stroke (AIS), and whether the PVS characteristics can predict later post‐stroke epilepsy (PSE). Methods A total of 312 patients with AIS were identified, of whom 58/312 (18.6%) developed PSE. Twenty healthy participants were included as the control group. The number of PVS in the basal ganglia (BG), centrum semiovale (CS), and midbrain (MB) was manually calculated on T2‐weighted MRI. The scores and asymmetry index (AI) of EPVS in each region were compared among the enrolled participants. Other potential risk factors for PSE were also analyzed, including NIHSS at admission and stroke etiologies. Results The EPVS scores were significantly higher in the bilateral BG and CS of AIS patients compared to those of the control group (both p < 0.01). No statistical differences in EPVS scores in BG, CS, and MB were obtained between the PSE group and the nonepilepsy AIS group (all p > 0.01). However, markedly different AI scores in CS were found between the PSE group and the nonepilepsy AIS group (p = 0.004). Multivariable analysis showed that high asymmetry index of EPVS (AI≥0.2) in CS was an independent predictor for PSE (OR = 3.7, 95% confidence interval 1.5–9.1, p = 0.004). Conclusions Asymmetric distribution of EPVS in CS may be an independent risk factor and a novel imaging biomarker for the development of PSE. Further studies to understand the mechanisms of this association and confirmation with larger patient populations are warranted.

followed up for at least 1 year post stroke or until the first occurrence of unprovoked seizures, whichever is sooner. This time point was chosen based on the observation that the incidence of late seizures is highest during the first year and typically peaks around 6-12 months after stroke. 1,19,20 Age-and sex-matched healthy volunteers were selected as controls. Only AIS patients and healthy controls who had undergone MRI brain with appropriate sequences were included for PVS analysis.
This study was approved by the medical ethics committee of the Nanjing Brain Hospital. All participants or their family members or authorized legal representatives provided written informed consent.

| Participants and groups
The inclusion criteria were as follows: (1) any type of AIS patient; (2) MRI obtained within 2 weeks from the stroke onset; and (3) over 16 years old. Individuals were excluded based on the following criteria: (1) history or family history of seizures or epilepsy; (2) history of mental illness, CNS infectious diseases, neurological immune disorders, metabolic disorders, febrile symptomatic seizures, and alcohol/drug withdrawal or intoxication; (3) previous neurosurgery, brain tumor, or traumatic brain injury; (4) patients who were not fit to receive or refused MRI scans at the stage of stroke onset; (5) brain MRI showing one of the following abnormalities: contusion, intracranial hemorrhage, neoplastic lesion, infectious or inflammatory lesion, or hydrocephalus; (6) quality of MRI too poor for evaluation (eg, due to motion artifact). Finally, 312 of the 1,465 patients met these criteria.
Post-stroke epilepsy was defined according to the ILAE diagnostic criteria for epilepsy 19,21 as the occurrence of at least one seizure ≥30 days after the stroke (which constitutes a high recurrence risk) or ≥2 seizures ≥7 days (late seizures) after the stroke, during the follow-up of the first year from AIS onset (PSE group). The latter definition of PSE reflects the operational definition of epilepsy from ILAE 21 and our clinical practice and was adopted in a previous study. 22 The clinical events were evidenced by medical records (must include EEG findings).
Acute ischemic stroke patients without seizures during the follow-up period were included in the no-epilepsy AIS group.
The screening resulted in the identification of 312 AIS patients based on the inclusion criteria. 20 healthy controls from the medical examination center of our hospital were identified, with the age range (55-80 years). The participants with any known neurological Conclusions: Asymmetric distribution of EPVS in CS may be an independent risk factor and a novel imaging biomarker for the development of PSE. Further studies to understand the mechanisms of this association and confirmation with larger patient populations are warranted.

K E Y W O R D S
asymmetric index, centrum semiovale, enlarged perivascular space, epileptogenesis, poststroke epilepsy condition or other diseases were excluded. This sample was necessarily much smaller than the AIS group due to the lower numbers of healthy individuals who are scanned at the hospital.

| PVS scores and asymmetry
Perivascular spaces were segmented manually by author N.Y, a neuroradiologist, with 5-year post-qualification experience. PVSs were assessed on T 2 -weighted MRI scans (resolution =1 × 1 × 6 mm, TR/TE/FA =7,411 ms/106 ms/90) acquired using a 3T MRI scanner (Siemens Verio) within 2 weeks after symptom onset of AIS. 23 The total sequences of T1, T2, DWI, and FLAIR were used to differentiate and quantify PVS. EPVSs were defined as tubular linear when parallel or round ovoid dot-like structures were perpendicular to the imaging plane with a CSF-like signal intensity (hyperintense on T2-weighted images) and a diameter of <3 mm. Different from EPVS, lacunar infarction is usually between 3 mm and about 15 mm in diameter and has a central CSF-like hypointensity with a surrounding rim of hyperintensity on FLAIR images, which EPVS lacks. 24 Figure 1 shows the examples of EPVS. For testing the inter-rater reliability of the method, MRI was initially reviewed by two trained raters (N.Y.and L.M.G. Pa radiologist with 2 years of experience) blinded to clinical details. The assessment of EPVS asymmetry was modified from previously described methods. 17,18 The difference between the right and left side at each location was calculated as an asymmetry index (AI): The total number of PVS of whole brain (S T ) was defined the sum of the EPVS values in the three locations: EPVS AI of the whole brain (AI T ) was calculated as follows: A higher AI value implies more asymmetric distribution of EPVS in the brain. As an unbalanced distribution of EPVS at some levels may be observed in healthy controls, 17,18 we used a threshold of AI ≥0.2 to define a high asymmetry in EPVS distribution, in accordance with Duncan et al 2018, 18 indicating that >60% of EPVSs were in one hemisphere. Based on this cutoff, an AI score of 0 (<0.2) and 1 (≥0.2) was assigned to each brain region.

| Statistical analysis
To measure inter-rater reliability, the second rater (L.

| Inter-rater agreement on EPVS numbers and scores
A total of 30 individuals from the AIS patients (n = 312) and healthy controls (n = 20) were evaluated by both raters to assess interrater reliability. Overall, there were moderate-to-good inter-rater

| Relevant factors for EPVS characteristics of AIS patients
In AIS patients, total EPVS numbers (BG, CS, and MB combined) increased with age (Pearson correlation, r = 0.22, p < 0.000). The scores and numbers of EPVS were not associated with stroke subtypes (duration time of symptoms before treatment, treatment methods after stroke, stroke causes stroke laterality, and infarct number) or severity (NIHSS at admission) (all p > 0.01).
The scores, numbers, and AI of EPVS at each observed region were not statistically different between in the unilateral stroke   Table 2, compared to the nonepilepsy AIS group, the PSE patients were younger, had higher NIHSS at admission, higher proportions of cortical lesions, and higher proportions of large-artery atherosclerosis and cardioembolic strokes.

| Clinical characteristics between AIS without epilepsy and with PSE
There were no differences in sex distribution, stroke laterality, and infarct numbers between the AIS without epilepsy and PSE patients.

| EPVS characteristics between the AIS without epilepsy and PSE patients
As shown in Table 3, there were no significant differences in EPVS scores of bilateral BG, CS, and MB between the PSE group and the nonepilepsy AIS group. There were also no significant differences on the total EPVS numbers and AI values between the PSE group and the nonepilepsy AIS group. The marked differences of AI score in the CS and midbrain region were found between the PSE group and the nonepilepsy AIS group (p = 0.004). There was no statistical difference of AI scores in the MB for the PSE and nonepilepsy AIS groups, though there was with the AI values. The AI score was not significantly different between groups in the other brain regions, or in the whole brain.

| A multivariate model for predicting PSE
Post-stroke epilepsy was predicted by a binary logistic regression model (Odds ratio 0.228, Nagelkerke R 2 = 0.518, p < 0.001) (   electrophysiologic imbalance of the brain, resulting in seizures. In line with this hypothesis, asymmetry of EEG background activity has been observed as an independent predictor of PSE during the first year after stroke. 26 Another potential explanation for the association between EPVS and PSE may be the relationship of the shared signaling pathways between epilepsy-related immunological inflammation in the brain (which could induce epilepsy and also be induced by epilepsy) and EPVSs. Growing experimental studies and clinical evidence have demonstrated that inflammatory reactions in the brain can increase the permeability of the blood-brain barrier (BBB) to proinflammatory molecules and cells and enhance neuronal excitability to trigger seizures. 27 Impaired PVS could allow leukocytes and antigen-presenting cells penetrate the glia, then releasing proinflammatory molecules further degrade BBB structures, which rarely occurs under basal conditions. 28 Furthermore, EPVSs are proposed to form part of a complex brain fluid drainage system to support interstitial fluid exchange and facilitate clearance of waste products from the brain. Impaired function of the PVSs may further lead to reduced blood flow, oxidative stress, hypoperfusion, and hypoxia, which are linked to PSE. 29 Moreover, Gaberel et al 30 have reported that the PVSs as glymphatic system were initially decreased and impaired around the infarct lesion in acute phase of 3 h after embolic ischemic stroke in mice, possibly due to its blocking effect of infarct lesion; however, this effect was reversed to the baseline level after 24 h of stroke, possibly due to its spontaneous arterial recanalization. In another experimental study, increased PVSs were observed at 48 h in a rat model of cerebral ischemia-reperfusion injury. 31 We also found this phenomenon of decreased PVSs around the infarct lesion in several cases of AIS within 24 h but totally with no statistical significance on the EPVS scores and counts between the infarct side and the noninfarct side. Maybe more rigors of study design on the onset phase were needed for this interesting question.
As many previous studies have reported, 1,3,4 our study also showed that cortical involvement, carotid circulation territory with large arteries, and stroke severity at admission are independent risk factors for PSE. However, there was no significant association between the predictors above and the EPVS-related parameters Our study has limitations. Firstly, not all the AIS patients were enrolled in the study, due to the limitation of MR acquirement at their acute stage. Secondly, the health control group was relatively smaller and younger, which was due to relatively fewer healthy people undergoing brain MRI. Thirdly, the other limitation for this study may be the method of counting EPVS manually, with potential subjectivity, although good inter-rater agreement was demonstrated. We also do not know whether and when the asymmetric distribution of EPVS would resolve over time, or whether it existed before the stroke. The latter would imply certain individuals may be intrinsically more prone to develop PSE. Lastly, the AI of EPVS did not contain the factor of PVS diameter. It is worth using the diameter of PVS to quantify asymmetry by some software in our future studies.
Our present findings suggest that asymmetric distribution of EPVSs in the CS may be a novel imaging biomarker for the development of PSE. An early predictor for PSE will provide better evidence and choice for early antiepileptic treatment. The AI score of EPVS provides a novel imaging biomarker for the understanding of epileptogenesis after stroke. Further studies should aim to explain the mechanism of this association and to confirm the association with larger patient populations.

| CON CLUS IONS
This study explored the factors influencing EPVS characteristics at the onset time of first AIS, and whether it could predict PSE.
The EPVS scores were significantly higher in the AIS patients compared to those of health controls. Higher asymmetry index scores of EPVS in CS were found between the PSE and nonepilepsy AIS groups. Interestingly, this asymmetric distribution of EPVS was not found in other brain regions between the PSE and nonepilepsy AIS groups. An asymmetric distribution of EPVS in CS may be an independent risk factor or a novel imaging biomarker for PSE.
Future studies should explore the relationship between EPVS and seizure types, seizure frequency, EEG findings, and AED treatment outcome.

CO N FLI C T S O F I NTE R E S T
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The authors confirm that the data supporting the findings of this study are available within the article after deidentification (text, tables, and figures). All the original data in statistics of this study can be got from the corresponding authors, upon reasonable request at any time. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines. The studies involving human participants were reviewed and approved by Nanjing Brain Hospital affiliated to Nanjing Medical University. All participants or their family members or authorized legal representatives provided written informed consent.