Functional connectivity alterations in patients with poststroke cognitive impairment: A resting-state functional magnetic resonance imaging study

Abstract

BACKGROUND

Cognitive impairment is a common functional impairment after stroke that severely affects the quality of life of patients. The underlying neurobiological mechanisms of poststroke cognitive impairment (PSCI) remain unclear.

AIM

To investigate the changes in functional connectivity (FC) in the brains of patients with PSCI.

METHODS

A total of 21 patients with PSCI and 12 healthy controls were selected as study subjects, and resting-state functional magnetic resonance imaging was performed. The brain region [Cerebellum_6_R (aal)] with significant differences identified by regional homogeneity analysis and the left thalamus, right thalamus, left basal ganglia, and right basal ganglia in the Brainnetome Atlas were selected as the seeds (regions of interest), and the FC between the seeds and whole-brain voxels was analyzed. Moreover, the 116 brain regions defined in the AAL116 atlas were selected as seeds (regions of interest), and the FC between the whole-brain seeds was calculated.

RESULTS

The results of the seed-based FC analysis revealed that the FC of the Cerebelum_9_R, Occipital_Mid_L, and Fusiform_R in the PSCI group was significantly greater than that in the control group. FC analysis of whole-brain seeds revealed that the FC of 20 pairs (Cerebelum_4_5_R and Cerebelum_6_R, etc.) in the PSCI group was significantly greater than that in the healthy control group.

CONCLUSION

Patients with PSCI exhibit changes in the FC of specific brain regions in the resting state, which may help researchers explore the underlying neurobiological mechanisms of PSCI from a new perspective.

Key Words: Functional connectivity; Ischemic stroke; Cognitive impairment; Resting-state functional magnetic resonance imaging; Stroke

Core Tip: Cognitive impairment is a common functional impairment after stroke that seriously affects the quality of life of patients. The underlying neurobiological mechanisms of poststroke cognitive impairment (PSCI) remain unclear. This study aimed to investigate the changes in brain functional connectivity of patients with PSCI, which may help explore the neurobiological mechanisms of PSCI from a new perspective.

INTRODUCTION

Cognitive impairment (CI) is a common form of functional impairment after stroke[1]. If not effectively managed, poststroke CI (PSCI) can result in abnormal mental behaviors and may even progress to rapidly advancing dementia[2-4]. However, the specific neurobiological mechanisms that cause PSCI remain unclear[5,6]. Images of brain activation obtained via blood-oxygenation-level-dependent functional magnetic resonance imaging (fMRI) based on the hemodynamic changes in different functionally active areas of the brain[7] have been widely used to study brain function in patients with neurological and psychiatric diseases[8]. Resting-state fMRI (rs-fMRI) measures the time correlation of changes in blood-oxygenation-level-dependent signals in various brain regions in the resting state[9]. Spontaneous neuronal activity in brain regions at baseline is detected through magnetic resonance imaging (MRI) scans, and the network connectivity of each relevant brain region is determined, reflecting spontaneous functional activity in the basal state[10]. Functional connectivity (FC), a commonly used method to assess brain function, is a term that refers to the level of functional connections between areas and is calculated by analyzing the correlation between one region and the rest of the brain[11]. FC can be used to quantify the temporal correlation of neurophysiological events in different brain regions and reveal the functional interactions between damaged brain regions and other brain regions, thus playing a critical role in the elucidation of the neurobiological mechanisms of PSCI[12,13]. Research on changes in FC in patients with PSCI may help researchers explore the underlying mechanisms of PSCI from a new perspective[6]. However, few studies have focused on changes in brain function in these patients. Thus, this study investigated changes in FC in patients with PSCI.

MATERIALS AND METHODS

Clinical data

T1-weighted structural imaging and rs-fMRI scans were performed on 21 patients with PSCI before treatment and 12 age- and sex-matched healthy controls (HCs) with a 32-channel head-neck coil (MR750 3.0 T, GE). T1W images were parameterized as follows: Repetition time = 8.2 milliseconds, echo time = 3.1 milliseconds, flip angle = 7°, matrix = 256 × 256, field of view = 256 × 256 (mm²), and slice thickness = 1 mm. The parameters of the rs-fMRI protocol were set as follows: Repetition time = 2000 milliseconds, echo time = 30 milliseconds, flip angle = 90°, matrix = 64 × 64, field of view = 230 × 230 (mm²), and slice thickness = 3.5 mm[14]. The demographic characteristics and clinical characteristics of the stroke patients, including the time since the stroke, lesion locations and volumes visualized on structural images, and detailed vascular risk factors, were systematically recorded. The lesions were located in the left basal ganglia and/or corona radiata in 8 patients, in the left thalamus in 1 patient, in the right basal ganglia and/or corona radiata in 9 patients, and in the right thalamus in 3 patients (Table 1). Cognitive ability was assessed using the Mini-Mental State Examination (MMSE) and Montreal Cognitive Assessment scores.

Table 1 The baseline demographic and clinical characteristics of the participants in this study.

Clinical characteristics	PSCI	HC	χ2/t/Z	P value
Sex (males/females)	12/9	4/8	1.733	0.188
Age (years)	60.2 ± 9.2	56.8 ± 5.0	1.393	0.174
Education level (years)	6.0 (6.0, 9.0)	6.0 (6.0, 9.0)	-0.514	0.645
Lesion laterality (left/right)	9/12	-	-	-
Onset time (days)	3.0 (2.3, 4.0)	-	-	-
Lesion volume (mm3)	1528.1 (630.5, 3055.3)	-	-	-
Hypertension (yes/no)	13/8	-	-	-
Diabetes (yes/no)	4/17	-	-	-
Smoking history (yes/no)	9/12	-	-	-
Alcoholism history (yes/no)	4/17	-	-	-

PSCI: Poststroke cognitive impairment; HC: Healthy control.

MRI data preprocessing pipeline

MRI data were preprocessed using the SPM12 and DPARSF5.4 toolkits in MATLAB 2022b[14,15]. The main preprocessing pipelines were as follows: (1) The first 10 time points were removed to ensure signal stability; (2) Time correction and head-motion correction were performed by excluding subjects whose head rotated more than 3° or moved more than 3 mm along the x, y, and z axes; (3) Spatial normalization of the MRI images of the subjects to the standard Montreal Neurological Institute space with a 3 mm × 3 mm × 3 mm resample was performed; (4) Spatial smoothing using a Gaussian kernel of a 6 mm × 6 mm × 6 mm full width at half maximum; (5) Linear drift was removed to eliminate the baseline drift caused by the machine; (6) Covariates were removed to eliminate the influences of head motion, white matter and cerebrospinal fluid signals; and (7) Low-frequency filtering was performed on all images by processing them with a horizontal bandwidth of 0.01-0.08 Hz to remove the high-frequency signals.

FC analysis

Within the MATLAB 2022b platform, the SPM12 and DPARSF5.4 toolkits were used for data preprocessing[14,15]. The FC analysis was conducted as previously described[11]. The significant brain regions [Cerebellum_6_R (aal)] identified in the regional homogeneity analysis and the left thalamus[14], right thalamus, left basal ganglia, and right basal ganglia in the Brainnetome Atlas were used as the seeds (regions of interest); the FC between the seeds and whole-brain voxels was analyzed. In addition, the 116 brain regions defined in the AAL116 atlas were selected as seeds (regions of interests), and the FC between the whole-brain seeds was calculated.

Statistical analysis

SPM12 was used for the analysis of FC between seeds and whole-brain voxels, and the GRETNA toolkit was used for the analysis of FC between whole-brain seeds. Paired t tests were performed on data from the PSCI group, with no covariates. Two-sample t tests were performed to compare the results between the PSCI group and the control group, and sex and age were used as covariates. For the analysis of FC between seeds and whole-brain voxels, AAL_116_binary_mask.nii was used as the statistical mask, with voxel_P < 0.001, cluster_P < 0.05, and FWE correction. For the analysis of FC between whole-brain seeds, the significance threshold was set to edge_P < 0.001, component_P < 0.05, and network-based statistics (NBS) correction according to previous studies[16-20].

RESULTS

Results of the analysis of FC between seeds and whole-brain voxels

An analysis of FC between seeds and whole-brain voxels showed that the FC of the Cerebelum_9_R, Occipital_Mid_L, and Fusiform_R in the PSCI group was significantly greater than that in the control group (Figure 1, Table 2).

Figure 1 Images of the brain regions in which functional connectivity between the seeds and the whole-brain voxels were analyzed (sectional view). A and B: Cerebelum_9_R driven by the brain region identified in the regional homogeneity analysis; C and D: Occipital_Mid_L driven by the left basal ganglia seed; E and F: Fusiform_R driven by the left thalamus.

Table 2 Functional connectivity analysis results between seeds and whole-brain voxels in the poststroke cognitive impairment group and the control group.

Seeds	Direction	Brain region	Coordinates of MNI peak			t value	Cluster size
			X	Y	Z
Brain region identified in the regional homogeneity analysis	PSCI group > control group	Cerebelum_9_R (aal)	15	-51	-45	5.4862	644
Left basal ganglia	PSCI group > control group	Occipital_Mid_L (aal)	-33	-72	27	4.4355	154
Left thalamus	PSCI group > control group	Fusiform_R (aal)	33	-60	-9	5.7931	221

MNI: Montreal Neurological Institute; PSCI: Poststroke cognitive impairment.

Results for the FC of whole-brain seeds

An analysis of the FC of whole-brain seeds revealed that the FC of 20 pairs in the PSCI group was significantly greater than that in the HC group. No decrease in FC was observed between the PSCI group and the control group (Figure 2, Table 3).

Figure 2 Analysis of the functional connectivity of whole-brain seeds showing the connectivity of brain regions with differences between the poststroke cognitive impairment group and the control group.

Table 3 Comparison of functional connectivity between the poststroke cognitive impairment group and the control group.

Inter-regional FC	PSCI group	Control group	t value	P value
Cerebelum_4_5_R and Cerebelum_6_R	1.06 ± 0.26	0.58 ± 0.22	4.5326	0.0001
Cerebelum_4_5_R and Cerebelum_8_R	0.68 ± 0.25	0.29 ± 0.24	4.0434	0.0004
Amygdala_L and Cerebelum_9_L	0.37 ± 0.21	0.06 ± 0.26	3.8097	0.0007
Cerebelum_4_5_L and Cerebelum_9_L	0.65 ± 0.20	0.24 ± 0.25	4.1890	0.0002
Cerebelum_4_5_R and Cerebelum_9_L	0.65 ± 0.17	0.19 ± 0.26	4.7038	0.0001
Cerebelum_6_R and Cerebelum_9_L	0.73 ± 0.32	0.33 ± 0.23	3.6753	0.0010
Cerebelum_4_5_L and Cerebelum_9_R	0.61 ± 0.25	0.20 ± 0.28	3.8625	0.0006
Cerebelum_4_5_R and Cerebelum_9_R	0.68 ± 0.21	0.20 ± 0.30	4.0530	0.0003
Fusiform_R and Vermis_3	0.44 ± 0.26	0.17 ± 0.15	3.8251	0.0006
Cerebelum_4_5_R and Vermis_3	0.71 ± 0.23	0.35 ± 0.25	3.8691	0.0006
Cerebelum_6_R and Vermis_3	0.62 ± 0.28	0.24 ± 0.17	3.9236	0.0005
Cerebelum_8_R and Vermis_3	0.56 ± 0.17	0.08 ± 0.17	4.4017	0.0001
Cerebelum_9_L and Vermis_3	0.52 ± 0.17	0.08 ± 0.29	3.7043	0.0009
Cerebelum_9_R and Vermis_3	0.53 ± 0.16	0.06 ± 0.30	3.9206	0.0005
Cerebelum_10_L and Vermis_3	0.44 ± 0.26	0.01 ± 0.17	4.4302	0.0001
Cerebelum_4_5_R and Vermis_4_5	0.85 ± 0.22	0.54 ± 0.14	5.6876	0.0000
Cerebelum_4_5_R and Vermis_8	0.72 ± 0.33	0.16 ± 0.35	4.2980	0.0002
Cerebelum_4_5_L and Vermis_9	0.61 ± 0.29	0.17 ± 0.30	3.9113	0.0005
Cerebelum_4_5_R and Vermis_9	0.61 ± 0.33	0.13 ± 0.26	4.3557	0.0002
Cerebelum_6_R and Vermis_9	0.75 ± 0.36	0.36 ± 0.23	4.2176	0.0002

The two-sample t tests of the data from the two groups revealed a significant positive functional connectivity (t value significantly > 0; that is, the poststroke cognitive impairment group > the control group, and the inter-regional functional connectivity of the poststroke cognitive impairment group increased significantly in the following brain regions). FC: Functional connectivity; PSCI: Poststroke cognitive impairment.

Correlation analysis

In the PSCI group, the partial correlation analysis (controlling for age) revealed that among the significant brain regions, the FC values of the Fusiform_R were significantly correlated with the MMSE scores (r = 0.464; P = 0.039; without correction for multiple comparisons), but no other significant correlations were identified.

DISCUSSION

Previous studies have shown that PSCI is closely related to infarcted brain areas[21]. The basal ganglia[22], internal capsule, thalamus, corpus callosum[23], angular gyrus[24], cortex cingulate[25] and subfrontal cortical areas[26] are key brain regions involved in global PSCI. Patients with PSCI have disrupted low-degree rich-club organization, relatively preserved functional core networks, and decreased feeder and local connectivity in cognition-related networks[27]. In the present study, as the ischemic stroke lesions of patients were located mainly in the bilateral thalamus or basal ganglia, in addition to the significant brain regions identified in the regional homogeneity analysis, the bilateral thalamus and basal ganglia were also used as the seeds. The results revealed that the FC of the PSCI group in Cerebelum_9_R, Occipital_Mid_L, and Fusiform_R was significantly greater than that of the control group. Although a seed-based FC analysis has been shown to be reliable and to effectively identify the regions with the strongest FC with seeds[28], it is strongly dependent on the selection of seeds[29]. In the present study, we also analyzed the FC of whole-brain seeds to avoid the exclusion of function-related voxels in the seed selection process[28]. The results revealed that, compared with that in the HC group, the FC of 20 pairs in the PSCI group increased. These brain regions with increased FC were located mainly in cerebellar regions. The increased FC in this study might be driven by a potential compensatory or maladaptive mechanism after ischemic stroke[30-34].

In recent years, researchers have studied patterns of altered FC in patients with PSCI from different perspectives[6,35-37]. Yue et al[6] performed fMRI examinations on 17 patients with PSCI and 24 HCs and used independent component analysis combined with sliding-window and K-means clustering approaches to examine the FC of 11 resting-state networks. Their results revealed that in terms of static functional network connectivity, the PSCI group presented decreased dorsal default mode network-ventral mode network, ventral mode network-salience network, dorsal default mode network-higher visual network, auditory network-right executive control network, and auditory network-visuospatial network interactions. Li et al[36] used rs-fMRI to study the effects of repetitive transcranial magnetic stimulation on brain function in patients with PSCI and reported that, compared with the control group, the repetitive transcranial magnetic stimulation group presented increased FC between the left dorsolateral prefrontal cortex and the precuneus, the inferior temporal gyrus, the middle and inferior frontal gyri and the marginal gyrus and decreased FC between the left dorsolateral prefrontal cortex and the middle temporal gyrus and thalamus. Lu et al[37] investigated functional alterations in the nucleus basalis of Meynert (NBM) and its projections in patients with mild CI and assessed the effects of computerized cognitive training (CCT). They analyzed FC between the NBM and three correlated brain regions. Their findings demonstrated that CCT intervention elevated the FC between the NBM and the right putamen, which lends support to the neuroplastic potential of impaired brains and highlights the clinical value of CCT in patients with MC.

As a core node in brain networks underlying social cognition, the cerebellum collaborates with multiple other cerebral regions to exert critical effects on cognitive processing. Nevertheless, the specific mechanism through which this cerebellumcentered network coordinates to ameliorate CI remains largely unclear. A growing body of prior research has also emphasized a robust link between cerebellar function and the pathogenesis of CI[32,38-42]. Zhang et al[43] used rs-fMRI to examine the difference in the FC of the limbic system and the cerebellum between patients with and without PSCI and identified the intracerebellar brain regions that exhibited functional changes during the onset of CI; their results revealed that the functional connections between the Cerebelum_Crus2_R and Frontal_Mid_Orb_L brain regions decreased. The functional connections between the Cerebelum_Crus2_R and Hippocampus_L brain regions decreased. Tang et al[44] collected rs-fMRI data from three different groups (28 patients with Alzheimer’s disease, 26 patients with mild CI and 30 HCs) and defined the cerebellar lobes (Crus II and IX) as seed regions to assess differences in cortical-cerebellar connectivity within groups, and reported decreased FC between the cerebellum and the medial frontal gyrus in patients with Alzheimer’s disease. However, the FC between the right-sided lobule IX and the medial frontal gyrus was significantly increased in patients with mild CI, which may compensate for the impaired memory that is often observed in these patients. Hong et al[42] collected rs-fMRI data from thirty-six participants with subcortical chronic stroke and thirty-eight HCs and performed FC analysis with the bilateral cerebellar anterior lobe and cerebellar posterior lobe as seeds for each participant. They reported that the cerebellar anterior lobe showed increased FC with the prefrontal cortex, superior/inferior temporal gyrus, and lingual gyrus, while the cerebellar posterior lobe showed increased FC with the inferior parietal lobule, precuneus, and cingulum gyrus in stroke participants compared with HCs. Jung et al[32] explored changes in hippocampal FC following ischemic stroke using rs-fMRI. Thirty-three patients with CI after ischemic stroke and sixteen HCs were recruited. Their results showed that across all the hippocampal subfields, FC with the inferior parietal lobule was reduced in participants with stroke compared with HCs, and the FC of the hippocampal subfields with the cerebellum was increased. In addition, a few studies have reported that cerebellar–hippocampal interactions are associated with various cognitive functions and have indicated the importance of the cerebellum and cerebellar-hippocampal connections for cognitive tasks. The results of this study differ from those reported by Zhang et al[43] and are consistent with those reported by Hong et al[42].

The results of this study revealed a positive correlation between the FC values of the Fusiform_R and MMSE scores after controlling for age, suggesting that it may be specifically associated with PSCI disease status; other regions with increased FC did not significantly correlate with MMSE scores, which may be related to the complexity of compensation and other factors[45].

This study had several limitations: (1) The sample size was relatively small, which significantly reduced the statistical power and significantly increased the risk of both type I (false positive) and type II (false negative) errors; (2) The patients with PSCI had mild CI according to the Petersen criteria[46], and the differences in brain FC between patients with different degrees of CI were not analyzed; and (3) The control for multiple comparisons was inadequate. The whole-brain seed analysis used the AAL116 atlas, and the significance threshold was set to edge_P < 0.001 and component_P < 0.05 (NBS-corrected). While the NBS is appropriate, the initial uncorrected edge threshold of P < 0.001 for such a massive number of tests may still be too lenient. In future studies, the sample size should be increased[47,48], subgroups with different degrees of CI should be established for further validation, and the impact of the test used to correct for multiple comparisons on the results should be further investigated.

CONCLUSION

In summary, this study preliminarily revealed that in the resting state, patients with PSCI exhibited changes in FC in specific brain regions, which may be important mechanisms of PSCI and may be potential imaging biomarkers for the evaluation of CI in PSCI patients.