Investigating altered brain functional hubs and causal connectivity in coronary artery disease with cognitive impairment

Neuropsychological examinations

All participants in this study underwent cognitive assessments using both the MoCA and the Mini-Mental State Examination (MMSE) scales to evaluate their cognitive status. The MoCA scale is a comprehensive assessment tool that evaluates different cognitive functions. These functions include visuospatial/executive abilities, naming skills, attention span, language proficiency, abstraction capabilities, memory recall, and orientation to time and place. By assessing these areas, the MoCA scale provides a holistic view of an individual’s cognitive abilities and helps identify any potential CI or deficits. The assessment of participants was conducted in a quiet environment following standardized procedures. The maximum score for both scales is 30 points. A total score below 26 points on the MoCA assessment was considered indicative of poor cognitive function, while a total score below 24 points on the MMSE assessment was considered indicative of poor cognitive function.

Magnetic resonance imaging protocol

Whole-brain images were acquired at the Shandong Provincial Hospital Affiliated to Shandong First Medical University using a Siemens 3.0 T Prisma MR system and a 64-channel head coil for brain scanning. All participants were positioned inside the machine, with foam padding placed around their heads to minimize movement. They were instructed to remain still and keep their eyes closed during the imaging process.

During the resting-state fMRI scan, an echo-planar imaging sequence was used to capture blood oxygen level-dependent (BOLD) signals with the following parameters: repetition time (TR) = 2,000 ms, echo time (TE) = 30 ms, field of view (FOV) = 220 × 220 mm², in-plane resolution = 64 × 64, thickness/gap = 3.5/0 mm, flip angle = 90°, 33 axial slices. T1-weighted whole brain magnetization-prepared rapid gradient echo (MPRAGE) images were collected to capture anatomical details using the following parameters: TR = 2,530 ms, TE = 2.98 ms, TI = 1,100 ms, FOV = 256 × 256 mm², in-plane resolution = 256 × 256 mm², flip angle = 7°, and 192 axial slices. During the entire scan session, participants were instructed to maintain a relaxed state and refrain from thinking about anything in particular.

Data preprocessing

The DPABI toolbox, which is based on SPM12, was used to preprocess all images (Yan et al., 2016). The preprocessing was performed using the MATLAB 2020 platform. The preprocessing pipeline included the following steps: (1) The initial 10 functional volumes were excluded to ensure signal stabilization. (2) The interleaved slice acquisitions in the functional data were corrected. (3) Head motion correction to exclude images with head motion >0.5 mm or rotation >0.5°, (4) Registration of the functional images to the MPRAGE anatomical image using the Montreal Neurological Institute (MNI) template, followed by spatial normalization. (5) Linear detrending. (6) Spatial smoothing using a 6 mm × 6 mm × 6 mm Gaussian smoothing kernel, full-width half maximum. (7) Filtering: performed bandpass filtering of 0.01–0.08 was performed to remove low-frequency drifts and high-frequency noise. The preprocessing steps of DC did not include smoothing beforehand.

Calculation of DC

DC is a fundamental approach in assessing the significance of connections within a brain network and identifying key nodes within it Li et al., (2016). To perform this analysis, the following steps are typically taken: 1. Whole-brain voxel-wise functional connectivity analysis was conducted on the preprocessed images. 2. Calculate correlation matrix: The time courses of each voxel in the subjects are correlated with those of all other voxels, resulting in a correlation matrix. To obtain the whole-brain correlation matrix, the Pearson correlation coefficient was calculated between each voxel and the remaining voxels in the brain. 3. Calculate voxel-wise degree of connectivity: The voxel-wise degree of connectivity is determined by summing the significant connections at the individual level. A correlation threshold, such as 0.25, is set to define which connections are considered significant (Buckner et al., 2009). 4. Standardize correlation matrix: The individual correlation matrix is standardized using Fisher’s z-transform. 5. Spatial smoothing: A Gaussian kernel with a full width at half maximum (FWHM) of 6 mm was applied to perform spatial smoothing. 6. Group-level comparisons: Group-level comparisons of the degree centrality maps are conducted using a general linear model (GLM) analysis, often implemented in the DPABI toolbox. 7. Statistical analysis: Two-tailed two-sample t-tests are performed to observe the differences in degree centrality values between each pair of groups. The voxel-level and cluster-level threshold is typically set to p < 0.001. 8. Correction for multiple comparisons: AlphaSim correction was applied to the results to control for false positives. The post-correction threshold is typically set to p < 0.05, with a cluster size threshold of ≥19 voxels (Ledberg, Åkerman & Roland, 1998).

Dynamic casual modeling

Based on the findings of the DC analysis, brain regions exhibiting significant changes in DC are identified as regions of interests (ROIs). These ROIs are then analyzed to determine the directionality and strength of the functional connections between them. This involved extracting the time series data from these regions and assessing changes in directionality and connectivity strength using an optimized model. SpDCM analyses were performed using Statistical Parametric Mapping (SPM12, http://www.fil.ion.ucl.ac.uk/spm), a widely-used software package for neuroimaging data analysis.

The time series data of specific brain areas were averaged for each subject. This averaging process was based on the outcomes of the DC analysis. During the analysis of resting-state fMRI data, a generalized linear model (GLM) analysis was conducted in a fully connected mode. This mode enabled the identification of bi-directional connections between each pair of ROIs within each subject. Importantly, it is worth noting that no external input was introduced to the model throughout the analysis process. In this study, the EC between a specific set of ROIs was calculated using 2⁴ free parameters. These parameters were utilized to quantify the strength and directionality of the connections between the ROIs.

To determine the most appropriate dynamic causal model for the two groups, a post hoc model selection process was employed (Friston & Penny, 2011). This involved fitting the full model with all free parameters and conducting a “greedy search” resulting in 256 reduced models (Rosa, Friston & Penny, 2012). Subsequently, a post hoc model optimization process was performed using the posterior probability as a criterion to identify the most suitable model for each group. After identifying the optimized model, the parameters (DCM.Ep.A) were compared using one-sample t tests and two-sample t tests within each group in SPSS 20.0 software. A significance level of p < 0.05 was considered statistically significant. These rigorous statistical analyses provided insights into the most appropriate dynamic causal models for each group and highlighted the differences in parameter inference between and within the groups.

Statistical analysis

In the present study, statistical analysis was conducted using SPSS 25.0. A two-sample t test was performed on the continuous variables in the clinical data of the subjects, including age, years of education, MoCA and MMSE scores. Additionally, a chi-square test was employed for comparing categorical data such as gender, hypertension, diabetes, and dyslipidemia. The significance level was set at 0.05. To evaluate the correlation between the strength of effective connectivity and clinical features of CAD, multivariate linear regression was employed while adjusting for confounding factors such as age, sex, hypertension, diabetes, and dyslipidemia. Statistical significance was defined as a p-value less than 0.05.

Comparison of DC values between CAD and HC

After applying appropriate adjustments, the statistical analysis revealed significant differences in the DC values of the left parahippocampal cortex (PHC), the right superior frontal gyrus (SFG), and the left medial temporal gyrus (MTG) between the two groups (Fig. 1). Specifically, compared to the HC group, the CAD group had increased DC values in the right SFG and decreased DC values in the left PHC and left MTG. The results can be found in Table 2.

Dynamic causal modeling analysis

In Figs. 2 and 3, our analysis has uncovered notable disparities in the strength of effective connectivity between individuals diagnosed with CAD patients and those in the HC. In comparison to the control group, the positive connectivity strength between the PHC and SFG, as well as between the PHC and MTG, was significantly reduced (p < 0.05) in CAD patients. At the same time, we observed an enhancement of negative connectivity between the PHC and SFG. Furthermore, we found that the positive connectivity strength between the MTG and SFG disappeared in CAD patients, and the positive connectivity strength between the SFG and PHC was not detected.

To explore the correlation between patients’ clinical scale and properties of effective connectivity, we conducted a Pearson’s correlation analysis between the MoCA scores, MMSE scores and the strength of effective connectivity. This analysis aimed to shed light on the potential relationship between cognitive abilities, as assessed by the MoCA and MMSE, and the communication patterns within the brain networks. By examining the strength of effective connectivity, we can uncover potential associations between cognitive performance and the functional connectivity of the brain. Notably, we observed a significant positive correlation between the strength of connectivity from the PHC to the MTG and the MoCA score in CAD patients (P < 0.05). This finding suggests that stronger connectivity between the PHC and MTG may be associated with better cognitive performance in CAD patients (Fig. 3B).

Differences in connectivity between the two groups

Our study differs from previous research in that we used spDCM to evaluate the differences in effective brain connectivity between two groups based on DC, thereby combining the strengths of both approaches. This enabled us not only to assess the static characteristics of brain networks and the importance of each brain node between the two groups but also to establish dynamic causal models between different brain regions, providing a more comprehensive analysis of the relationships between brain regions. As academic researchers, we believe that this approach offers a more sophisticated and nuanced understanding of differences in brain connectivity. In our study, we utilized resting-state fMRI analysis with spDCM to investigate connectivity patterns in patients with CAD compared to controls. By examining the strength of connections, interpreted as rate constants of neuronal responses, we aimed to identify specific effective connectivity and their directionality in the resting-state fMRI data.

The PHC is a very important region in the brain, which is a critical structure for the function of the hippocampus and plays a role in higher cognitive functions. The connections between the PHC and other brain regions are important for overall cognitive function, memory function, visual recognition, and more (Eichenbaum, Yonelinas & Ranganath, 2007). Therefore, we can infer that when the effective connections between the PHC and other brain regions decrease, the balance between brain regions may be disrupted, leading to a decline in cognitive ability in CAD patients. Our study found that the information flow between the PHC and MTG in CAD patients was blocked, which may lead to impaired cognitive function. Milton et al. (2012) used fMRI to investigate brain activity during an autobiographical memory task in the transient epileptic amnesia (TEA) population. The study found that the effective connectivity between the PHC and the MTG was significantly decreased in the TEA group compared to the HC, indicating that a decrease in effective connectivity between the PHC and MTG may lead to a decline in memory and overall cognitive ability in patients (Milton et al., 2012). Xia, Fu & Qian (2016) used resting-state fMRI to investigate the role of PHC functional connectivity changes in memory impairment after concussion. The study found that the functional connectivity between the PHC and MTG was significantly decreased in patients with memory impairment after concussion compared to healthy controls, and the study also found that the degree of functional connectivity decrease was significantly correlated with the severity of memory impairment. In addition, the study found that the connectivity changes between the parahippocampal gyrus and the temporal lobe after concussion may be related to neural network reorganization in the brain. In our correlation analysis, we also found that the decrease in effective connectivity between the PHC and MTG in CAD patients was significantly correlated with the severity of CI. This indicates that the decrease in effective connectivity between the PHC and MTG has an important impact on the decline in cognitive function in CAD patients.

The SFG is a key node in both the default mode network (DMN) and the cognitive control network (CEN), and changes in its effective connectivity with other brain regions may significantly affect cognitive function. Our study showed that CAD patients had significantly lower effective connectivity strength between the SFG and PHC compared to HC, and the connection direction changed from positive in HC to negative in CAD patients. This suggests that SFG positively stimulates PHC to complete cognitive tasks differently in CAD patients compared to HC, and the connectivity between SFG and PHC is inhibited in CAD patients, which may be an important reason for the decline in cognitive function in CAD patients. Peng & Burwell (2021) explored the importance of PHC and SFG in regulating context-modulated behavior and emphasized the importance of transmission between SFG and PHC in context-modulated behavior, formation of memories, emotion regulation, and decision making. They proposed that the connection between SFG and PHC may be more important than the connection between the hippocampus and parahippocampal gyrus (Peng & Burwell, 2021). This suggests that the decrease in effective connectivity between SFG and PHC may have serious effects on cognitive function. In our results, we also found that the limited connectivity strength between MTG and SFG decreased in CAD patients, and the connection direction changed from positive in HC to negative in CAD patients. MTG is a crucial brain region, and Xu et al. (2019) employed resting-state fMRI and coactivation pattern analysis to investigate the functional characteristics of the temporal gyrus. By utilizing these techniques, they aimed to gain a deeper understanding of the temporal gyrus and its role in brain function. They found that the temporal gyrus plays different roles in language, memory, and social cognition, and there are multiple subregions within the temporal gyrus that are closely related to cognitive functions, such as functional connectivity (Xu et al., 2019). In terms of neural fiber connectivity, Briggs et al. (2021) used diffusion imaging to study the potential fiber bundle structure of MTG. The results showed that MTG is highly involved in the brain’s language networks and higher-order cerebral networks such as the DMN (Briggs et al., 2021). In terms of brain network connectivity, Lu et al. (2022) found that the functional connection between the MTG and the SFG is closely related to individual thought inhibition, emphasizing the importance of the SFG and the MTG. The functional connection between these two regions may play an important role in controlling negative emotions and inhibiting thoughts (Lu et al., 2022). These studies, similar to our results, reveal the neuroimaging mechanism underlying changes in cognitive function by measuring changes in effective connectivity between different brain regions in CAD patients. They highlight the important role that changes in effective connectivity strength and direction between the SFG, PHC, and MTG can play in altering cognitive function.

Limitation

Our study does have certain limitations that should be acknowledged. One of the main limitations is the relatively small sample size, which consisted of only 24 patients with CAD and 24 HC participants. The small sample size may limit the generalizability of our findings and reduce the statistical power of our analyses. There may be limitations in exploring the effective connectivity of brain networks between the two groups, and our study subjects were recruited from a single medical institution, which may have recruiting bias. For example, the study may not have included all populations with CAD. Second, in terms of neuropsychological examination, we only tested the MMSE and MoCA scales for the subjects and evaluated the cognitive function of the subjects based on the scores of the two scales. In the future, we will increase the number of scales to more specifically evaluate the degree of decline in various cognitive functions. At the same time, this study only considered the differences in network connections between CAD patients and HC, but there may be other potential factors that could affect the research results, such as drug treatments and individual differences among patients. In future studies, we will further increase the number of subjects and optimize the experimental process in order to achieve more accurate experimental results and more discoveries.