This was a single-center, retrospective, observational study. The clinical and neuroimaging data were obtained from a prospectively collected cohort (Registry for Critical Care of Acute Stroke in China, registered at www.chictr.org.cn, identifier ChiCTR1900022154). Patients with large vessel occlusive (LVO) stroke, which included the middle cerebral artery (MCA), who had undergone EVT in Beijing Tiantan Hospital, aged 18–85 years, and premorbid modified Rankin Scale (mRS) 0–2, were eligible to enroll. We excluded study participants with infarcts within vertebrobasilar artery territory, those with bilateral hemisphere infarcts, and those with severe systemic medical comorbidities or terminal illness with an anticipated life expectancy of less than one year (including conditions such as congestive heart failure [New York Heart Association Class IV], hepatic failure, and malignant tumors). We also excluded those enrolled in other clinical trials, those with poor acoustic temporal windows, and patients who could not tolerate the EEG-TCD monitoring due to delirium, or those requiring sedation or anesthesia during data acquisition. Management decisions adhered to the latest relevant guidelines available at the time of patient enrollment [13, 14]. The Ethics Committee of Beijing Tiantan Hospital approved the study. Written informed consent was obtained from the legally authorized representatives for all individual participants included in the study.

Patients were monitored synchronously with EEG (Nicolet, Natus Medical, USA), TCD (DWL, Compumedics DWL, Germany), and arterial blood pressure in the neurocritical care unit setting within 48 h after EVT. The patients lay quietly in bed, and no stimulation or task was applied during the monitoring. EEG electrodes were placed on the scalp according to the international 10–20 system. To represent the cerebral regions supplied by MCAs with fewer electrodes, three pairs of bipolar channels (F3-C3/F4-C4, T3-P3/T4-P4, and P3-O1/P4-O2) were selected for analyzing the EEG signal. Two 2.5-MHz TCD transducers, fitted on a headband, were placed over the temporal window. The depth of insonation was adjusted between 50 and 65 mm to measure the CBFV of bilateral MCA (Fig. 1a).

Data acquisition and contralateral neurovascular coupling. a We synchronously recorded the CBFV of MCA and the EEG signals from the brain region supplied by MCA. The TCD probes were placed at the temporal windows, with a measurement depth of 50–65 mm, reflecting the M1 segment of MCA. We selected three pairs of bipolar channels (F3-C3/F4-C4, T3-P3/T4-P4, and P3-O1/P4-O2) to cover the MCA territory (red area). The ACA and PCA territory were represented by yellow and green areas, respectively. b PACFC was used to bridge the EEG from the stroke-side hemisphere and CBFV of the contralesional MCA to calculate the contralateral NVC. Bilateral A1 segments of ACA (③) and ACoA (④) forms the aCoW, which serves as the primary collateral connecting bilateral ICA (①) and MCA (②). Of note, the aCoW is not always complete. As shown in the figure, cerebral blood flow signal was recorded in the M1 segment of MCA located at the distal to ICA. Therefore, the contralateral NVC in this study was unlikely to represent the primary collateral through the aCoW (blue pathway), but rather reflected collaterals via the leptomeningeal and other anastomoses (green pathway). ACA, anterior cerebral artery, ACoA, anterior communicating artery, aCOW, anterior Circle of Willis, CBFV, cerebral blood flow velocity, EEG, electroencephalography, ICA, internal carotid artery, MCA, middle cerebral artery, NVC, neurovascular coupling, PACFC, phase-amplitude cross-frequency coupling, PCA, posterior cerebral artery (Color figure online).

Phase-amplitude cross-frequency coupling (PACFC) was used to assess the contralateral coupling between EEG amplitude from the stroke hemisphere and CBFV of the contralesional MCA (Fig. 1b). First, artifact-free data were manually selected, and CBFV was filtered (0.05–0.15 Hz) to reduce respiratory-related fluctuations. Using Hilbert transform, we extracted EEG amplitude and CBFV phase. The modulation index was calculated to quantify coupling, with high modulation index indicating strong neurovascular coupling. Surrogate data generated a null distribution, fitting a Gaussian to obtain the mean (μ) and standard deviation (σ). Normalized PACFC was computed for statistical analysis. Fifty 5-min data segments were analyzed per participant, considering EEG frequency bands (delta, theta, alpha, beta) and other metrics. The detailed process of calculation was presented in our previous work [11] and the supplementary materials.

To better understand the differences in PACFC values across frequency bands, and exclude possible confounding factors, we also calculated the metrics including ipsilateral PACFC values, alpha to delta power ratio (ADR), and mean velocity index (Mx) (a cerebral autoregulation parameter) on stroke and contralesional sides [15]. All the computations were performed using Matlab 2022b (MathWorks, Massachusetts, USA) and Python 3.12.

The National Institute of Health Stroke Scale (NIHSS) score at admission to the hospital and at 7 days (or at discharge from the neurocritical care unit, if earlier than 7 days), along with mRS scores at 90 ± 7 days were collected by investigators independent of the EEG-TCD monitoring. We defined ΔNIHSS (ΔNIHSS = NIHSS scores at 7 days − NIHSS scores at admission) to represent early neurological function changes. mRS score 0–2 at 90 days was regarded as a favorable outcome.

All patients underwent computed tomography (CT) angiography and perfusion imaging, or MRI scanning, at baseline and 24 h after the EVT procedure. A plain CT was performed at 7 days and whenever necessary thereafter. The infarct size before the procedure and at 7 days were estimated by RAPID software (iSchemaView, California, USA) and manual measurement by independent neuroimaging raters, respectively. Infarct growth was defined as the difference between the infarct volume at 7 days and the baseline.

To investigate how the completeness of the anterior Circle of Willis (aCoW) affects the contralateral coupling, we further performed subgroup analysis by dichotomizing the study participants into complete and incomplete aCoW subgroups. The aCoW was incomplete if the anterior communicating artery or A1 segment(s) of anterior cerebral artery (ACA) was hypoplastic or absent in either hemisphere [16] as determined by digital subtraction angiography or CT angiography.

Continuous variables were summarized as means ± standard deviation for normally distributed data, and medians with interquartile ranges (IQRs) for nonnormally distributed data. Categorical variables were reported in numbers (%). We used t-test, χ2 test, Fisher’s exact test, and Wilcoxon rank-sum tests as appropriate for unadjusted comparisons. The Spearman correlation coefficient was applied to examine the correlation between the two groups. Odds ratios (ORs) and 95% confidence intervals (CIs) of dichotomous variables were estimated using a logistic regression model. We standardized the variables into normal distribution to avoid the unproportionally impact from variables with large magnitude on the regression model. In the multivariable regression, we accounted for known predictors of outcome after LVO stroke, including age, preprocedural NIHSS score, ASPECTS, 24-h stroke volume, and modified Thrombolysis in Cerebral Infarction (mTICI) score. Statistical analyses were performed using StataMP 17.0 (Stata Corporation, Texas, USA). P values < 0.05 were considered statistically significant.