Our study has a small-sample exploratory design. Between January 2022 to May 2023, 18 patients with symptomatic unilateral CICAO and who were candidates for endovascular recanalization were enrolled into the study. Confirmed symptomatic CICAO due to a diagnosis of transient ischemic attack (TIA) or infarction within the distribution territory of the affected carotid artery during 6 months preceding enrolment in the study. The following inclusion criteria were applied: (1) age $\ge$45 years; (2) completely occluded internal carotid artery (ICA) confirmed by digital subtraction angiography (DSA) and the presence of a stump with proximal ICA lumen patency. Additionally, the cavernous and/or petrous segments needed to exhibit reconstitution via collateral filling from branches of ipsilateral external carotid artery, anterior communicating artery (AcomA), and/or posterior communicating artery (PcomA). This corresponds to the Type A and Type B classifications of CICAO patients as defined by Hasan et al. [27]; (3) A modified Rankin score (mRS) of $\le$2. (4) education level ($\ge$6 years). The exclusion criteria included: (1) ipsilateral middle cerebral artery stenosis $\ge$50% or contralateral carotid artery or middle cerebral artery stenosis $\ge$50%; (2) non-atherosclerotic carotid artery occlusion, such as that caused by arteritis or dissection; (3) Language comprehension or expression disorders that prevent or hinder cooperation with neuropsychological assessment; (4) concurrent neurodegenerative or psychiatric diseases; (5) congestive heart failure [left ventricular ejection fraction (LVEF) 40%]; or severe renal failure [estimated glomerular filtration rate (eGFR) of 30 mL/min/1.73 m²]; or severe liver failure (Child-Pugh Class C); or malignancy; (6) contraindications for an MRI scan (metal implants); (7) acute carotid artery occlusion. In addition, 15 HCs without cerebral large artery stenosis or occlusion (confirmed by ultrasound) were recruited. Our study was conducted in compliance with the Declaration of Helsinki. Patients or their families/legal guardians provided written informed consent. The study protocol was approval by the Institutional Review Board of the First Affiliated Hospital of Zhejiang university.

Patients were administered two antiplatelet agents (aspirin and clopidogrel) for a minimum of 5 days prior to undergoing endovascular revascularization. The procedure was performed via the femoral artery access under local anaesthesia and general heparinization, utilizing embolic-protection devices. Technical success was characterised by the successful implantation of stents following the recanalization of the occluded carotid artery, achieving a final residual diameter stenosis of 20% or less and modified Thrombolysis in Cerebral Infarction (mTICI) grade 3 antegrade flow [28]. Post-recanalization systolic blood pressure was maintained between 100 and 140 mmHg. Patients were prescribed with antiplatelet agents (aspirin and clopidogrel) for a duration of three months post-intervention and followed by lifelong monoclonal antiplatelet therapy. The occurrence of neurological sequelae, intracranial haemorrhages, and mortality was documented in the postoperative period. Additionally, computer tomography (CT) angiography (CTA) was scheduled at the 6-month follow-up.

In our study, neuropsychological evaluations were conducted by a neurologist who had undergone specialized training. Mini-Mental State Examination (MMSE) (Chinese version) [29] and Montreal Cognitive Assessment (MoCA) (Beijing Version) [30] were used to assess the global cognition. Trail Making Test (TMT)-A and -B were used to examine visuospatial and executive function [31]. Symbol Digit Test (SDT) [32] was employed to examine visual search, perception, and graphomotor speed. Digit Span Test (DST), including Forward DST (FDST) and Backward DST (BDST) [32], was used to evaluate working memory. Neuropsychological assessments were conducted within 7 days before surgery for patients and at baseline for healthy controls. Identical assessments were repeated for patients at the 6-month follow-up after carotid artery recanalization.

Patients received brain MRI sessions at two time points: 1 week prior to carotid recanalization and 6 months following the procedure. Imaging data were obtained using 3.0-Tesla MR system (SIGNA Architect, General Electric (GE) Medical systems, Waukesha, WI, USA and DISCOVERY MR750, GE). A standard 19-channel head coil was used to receive the signal (GE Healthcare, serial number, SN: H01511, Aurora, OH, USA). The subjects were placed on supine, with their heads supported with foam pads and a belt. They were required to close their eyes, avoid thinking about anything and falling asleep during the MRI.

In the SIGNA Architect System (SN: PG75A1900119SC, GE Medical systems), T1 parameters were as follows: repetition time (TR) of 7.7 ms; echo time (TE) of 3.1 ms; field-of-view (FOV) of 256 mm; voxel size of 1.0 $\times$ 1.0 $\times$ 1.0 mm; slice thickness of 1.0 mm; 176 slices; and a total scanning time of 4 min and 39 s. The parameters of rs-fMRI sequence included: TR of 2000 ms; TE of 30.0 ms; voxel size of 3.4 $\times$ 3.4 $\times$ 3.6 mm; FOV of 220 mm; slice thickness of 3.6 mm; 36 slices (6 patients and 3 HCs); and total scanning time of 6 min and 40 s. In the DISCOVERY MR750 System (SN:PG75E1900019SC, GE Medical systems), T1 parameters included: TR of 8.2 ms; TE of 3.2 ms; voxel size of 1.0 $\times$ 1.0 $\times$ 1.0 mm; FOV of 256 mm; slice thickness of 1.0 mm; 180 slices; and total scanning time of 4 min and 52 s. The parameters of rs-fMRI scanning included: TR of 2000 ms; TE of 30.0 ms; voxel size of 3.4 $\times$ 3.4 $\times$ 3.2 mm; FOV of 220 mm; slice thickness of 3.2 mm; 45 slices (12 patients and 12 HCs); and total scanning time of 6 min and 40 s.

The initial pre-processing step included flipping the images of patients with left-sided carotid occlusion, standardizing the affected hemisphere to the right-hand side across all participants. Pre-processing of rs-fMRI and T1 data was conducted using the data processing & analysis for brain imaging (DPABI) package (Version 7.0, The R-fMRI Lab, Institute of Psychology, Chinese Academy of Sciences, Beijing, China) [33]. The first 10 time points were excluded to stabilize the magnetic field and allow participants to adapt. Slice timing was corrected by matching the centre slice, while head motion was corrected by spatial realignment to the mean volume of a series of images. The T1 images were co-registered to the fMRI image, segmented and normalized to a group specific template in Montreal Neurological Institute (MNI) space via the Diffeomorphic Anatomical Registration Through Exponentiated Lie Algebra (DARTEL) toolkit [34] within DPABI package (http://rfmri.org/DPABI). Detrending was applied to fMRI images to remove signal intensity drift. Nuisance regression was carried out to remove physiological noise from cerebrospinal fluid, white matter, and motion (Friston 24 method). The last pre-processing step included applying band-pass filtering between 0.01 and 0.10 Hz, followed by image smoothening using a 4-mm full-width half-maximum (FWHM) Gaussian kernel to reduce the registration error between subjects and improve signal-to-noise ratio.

After the pre-processing, seed-based voxel-wise functional connection analysis was conducted using DPABI. The following 6 networks were evaluated based on a previous study in patients with carotid artery stenosis [15]: salience network (SN), at the seed of dorsal anterior cingulate cortex (–1, 10, 46); visual network (VN), at the seed of left primary visual cortex (–4, –81, –10); default mode network (DMN), at the seed of posterior cingulate cortex (0, –50, 22); sensorimotor network (SMN), at the seed of left primary motor cortex (–41, –20, 62); frontoparietal network (FPN), at the seed of left middle frontal gyrus (–45, 29, 32); and dorsal attention network (DAN), at the seed of left frontal eye field (–26, 6, 48). All spherical seeds with a 4 mm radius were positioned contralateral to the occlusion side.

R version 4.4.1 (R Foundation for Statistical Computing, Vienna, Austria) was utilized to analyse data in the two groups. Data normality distribution was tested by the Shapiro-Wilks normality test. Those that followed a normal distribution were presented as the mean $\pm$ standard deviation (SD), and a t-test was used to analyse inter-group comparison. Non-normally distributed data was expressed as a median (interquartile range (IQR)), and Mann-Whitney U test was used for inter-group comparison. Fisher’s exact test was used to compare categorical variables between the two groups. A two-tailed significance level of p 0.05 was applied in all comparisons.

At the image level, one-sample t-test were performed for both the control and the patient groups before and after endovascular recanalization, with a minimum cluster size of 40 voxels and a threshold of p 0.001. A two-sample t-test was performed to compare FC values within each of the six networks between patients and healthy controls, with the gender, age, instrument model and Mean framewise displacement (FD) Jenkinson as covariates [35]. A paired t-test was assessed the pre- and post-operation FC strengths in the patient group with the mean FD Jenkinson as covariates. The Gaussian random field (GRF) correction was applied with the voxel level p 0.005 and cluster level two-sided p 0.05.

To explore the relationship between FC alterations and baseline cognitive performance, the average of the FC strengths for each patient (N = 18) were extracted from voxels in each distinct cluster. Pearson correlation analysis was conducted to associate these FC values with their respective cognitive scores. The relationship between FC alterations in distinct clusters and the changes in cognitive assessment pre- and post-operation was analysed using partial correlation analysis, with education years and age included as covariates. All Pearson correlation analysis were two-sided, and a p 0.05 was set as a cut-off for significance.

Prior to rs-fMRI analysis, we established the head movement threshold (3.0 mm and 3.0 degree in max head motion) [33] to exclude subjects with excessive head motion. Ultimately, no patients were excluded based on this criterion. Among the 18 patients, infarct lesions were predominantly located in watershed areas, with 15 cases in the internal watershed, 2 cases involving both anterior and posterior watershed regions, and 1 case in the anterior watershed region. The Circle of Willis was evaluated via DSA. The AcomA was observed in eight patients, the PcomA in six patients, with AcomA and PcomA observed in four patients. All patients with symptomatic CICAO underwent endovascular recanalization procedures. Due to the inability of a micro-guidewire to enter the true lumen, there were two failed cases of endovascular recanalization. Another patient who was successfully recanalized suffered a transient exacerbation of neurological symptoms due to new ischemic infarcts in the hemisphere ipsilateral to occlusion which confirmed by MRI scan. No other patients suffered any perioperative complications. Therefore, after excluding the aforementioned three cases, 15 patients completed the follow-up.

The baseline characteristics and cognitive function evaluations are presented in Table 1. Comparative analysis confirmed that the two groups did not markedly differ in age, sex, education, and vascular risk factors. Marked differences were observed in cognitive performance, with patients with symptomatic CICAO exhibiting lower MMSE, MoCA, SDT, FDST, and BDST scores. Conversely, patients presented with significantly higher TMT-A and -B scores. These findings suggested a decline in general cognition and several specific cognitive domains, including attention, working memory, visual-spatial function and executive function in the symptomatic CICAO patients compared to the HCs.

The CTA performed at 6 months in 15 patients and confirmed no restenosis or occlusion. During the 6-month follow-up period, patients who achieved successful recanalization had no recurrent stroke or TIA. The MMSE, MoCA, and FDST scores were higher in the patient group after carotid recanalization (Table 2).

In the patient group before endovascular recanalization, the side with ipsilateral to the occlusion (right) exhibited asymmetrical hypo-connectivity within the 6 selected networks, with seeding performed at the occlusion contralateral side (left; Fig. 1). Hypo-connectivity of a specific network in patients with symptomatic CICAO was defined by abnormally lower connectivity strengths compared to those observed in HCs. At baseline, for the DAN seed, the patients group showed disrupted FC particularly within the ipsilateral (occlusion side, right) frontal pole and middle frontal gyrus compared to the HCs. The DMN seed region revealed reduced FC with the ipsilateral angular gyrus, while the VN seed region showed reduced FC with the ipsilateral fusiform gyrus (Fig. 2, Table 3). Additionally, the FPN seed region displayed reduced FC with the middle temporal gyrus on the side contralateral to the occlusion (Fig. 2, Table 3). Notably, the between-group differences in the SMN and SN were not significant.

Fig. 1.

A within-group examination of six brain networks in patients and healthy controls. The predefined regions of interest (ROIs) for specific networks are indicated by hollow circles. Patients with left-sided of carotid occlusion were flipped to the right side. Color bars indicate T scores. DAN, dorsal attention network; SMN, sensorimotor network; FPN, frontoparietal network; DMN, default mode network; VN, visual network; SN, salience network.

Fig. 2.

Group comparisons of six brain networks between patients and HCs. Clusters with significant increments of functional connectivity are shown in red. Patients with left-sided of carotid occlusion were flipped to the right side. Color bars indicate T scores (voxel-level p 0.005, cluster-level p 0.05).

We further assessed the correlation between FC changes and cognitive function, revealed: a positively correlation between a reduction in DAN with frontal pole (ipsilateral to occlusion) and the total score of MMSE (r = 0.499, p = 0.049), MoCA (r = 0.515, p = 0.041), BDST (r = 0.594, p = 0.015); and a negative association with TMT-A (r = –0.563, p = 0.023) and TMT-B scores (r = –0.602, p = 0.014) (Fig. 3).

Fig. 3.

Pearson correlation analysis between FC strength and neurocognitive performance at baseline. DAN1, dorsal attention network (at the frontal pole ipsilateral to occlusion); BDST, backward Digit Span Test.

Six months after endovascular recanalization, the inter-hemispheric FC exhibited a symmetrical shape, particularly in the DMN, similar to manifestations observed in HCs (Fig. 4). The two-sample paired t-test, adjusted for mean FD of head motion, revealed significant FC increases. Specifically, the DMN seed region exhibited enhanced FC with the middle occipital gyrus ipsilateral to occlusion, and the VN seed region showed increased FC with the fusiform gyrus ipsilateral to occlusion following endovascular recanalization (voxel-level p 0.005 and cluster-level p 0.05, corrected by GRF) (Fig. 5).

Fig. 4.

FC maps at the group-level in the DMN. From left to right: healthy controls (A), CICAO patients before carotid recanalization (B), 6 months after carotid recanalization (C). Color bars indicate T scores (voxel-level p 0.001; cluster $>$40 voxels). FC, functional connectivity.

Fig. 5.

Seed-to-voxel contrast between post-operation and pre-operation of carotid recanalization showed significantly increased FC strengths in DMN with right Middle Occipital Gyrus (A) and VN with right Occipital Fusiform Gyrus (B). Patients with left-sided carotid occlusion were flipped to the right side. Color bars indicate T scores (voxel-level p 0.005).

The improvement scores for the patient group were calculated by subtracting the post-operative MMSE, MoCA, SDT, TMT-A, TMT-B, FDST and BDST, from preoperative the scores, and then taking the absolute values. The mean FC values for preoperative and postoperative FC were extracted for two distinct cluster, DMN and VN, and partial correlation analysis revealed a positive association of the preoperative FC values of the DMN with the improvement in TMT-A score (r = 0.629, p = 0.021) (Fig. 6).

Fig. 6.

Partial correlation analysis uncovered a positive association between the preoperative FC of DMN and improvements in TMT-A score.