Increased Effective Connectivity of the Left Parietal Lobe During Walking Tasks in Parkinson’s Disease

INTRODUCTION

Individuals with Parkinson’s disease (PD) experience gait disorders that result in an increased risk of falling and a substantially decreased life quality [1]. Early gait disorders are characterized by reduced gait speed and stride length and increased gait variability [2, 3], suggesting that changes in walking capabilities in PD are associated with decreased gait automaticity [4]. We noted that gait automaticity involves cortical and subcortical structures [5, 6]. In healthy people, when the movement is automated, brain functional activity in many brain regions—such as the prefrontal lobe, dorsolateral cortex, anterior cingulate gyrus, anterior supplementary motor area, anterior motor area, parietal lobe, and cerebellum—decreases. Simultaneously, the functional connectivity between motor-related brain regions (motor cortex, basal ganglia, cerebellum) is enhanced, and the functional connectivity between motor- and attention-related brain regions is decreased [7], suggesting that the realization of motor automation involves both cortical and subcortical structures. Healthy people can maintain gait consistency quasi-autonomously. Therefore, subcortical striatum dysfunction in individuals with PD leads to impaired gait automation and increased gait variability. The walking process requires either cortical or cognitive control and continuous attention, which may involve compensation for cortical function [6]. Therefore, it is important to obtain a better understanding of the patterns of interaction between different neural regions during walking.

Functional brain imaging has been used to investigate the patterns of interaction between neural regions. Functional connectivity (FC) and effective connectivity (EC) are commonly used concepts when evaluating functional brain imaging. FC is thought to measure temporal correlations and dependence between spatially separated brain regions [8, 9], thus revealing synergism between different brain regions, but it cannot support inferences regarding directed brain connections [8]. In contrast to nondirected FC, EC refers to the impact that one neural system has on another [10]. EC has provided a greater level of information regarding the communication of information across different neural regions and offers a better understanding of the models of interaction between these regions [11–13].

Prior studies have reported EC patterns in PD based on electroencephalogram (EEG), functional magnetic resonance imaging (fMRI), and positron emission tomography (PET) findings [14–16]. However, a major limitation of these studies is that the participants could not walk on a pathway in a realistic environment. Most research utilized basic, single-segment motor tasks (e.g., fingertip tapping or button pushing) or task simulations (e.g., virtual reality or guided imagery) to infer cortical or subcortical activity related to general motor control [17]. There has been little research of EC patterns in the brain cortex in PD during walking tasks.

Functional near-infrared spectroscopy (fNIRS) is an optical neuroimaging technique used to measure local hemodynamic changes in the cerebral cortex [18]. Compared with EEG, fMRI, and PET, portable fNIRS allows the recording of brain signals during walking tasks in real environments. A meta-analysis of PD based on fNIRS including 10 studies that focused on the oxyhemoglobin in the prefrontal cortex (PFC) of individuals with PD during different difficulty levels of ambulatory walking tasks showed that, compared to healthy older individuals, individuals with PD in the on state showed increased mobilization of the PFC, while during the off-state no significant changes were identified [19]. In another review, which included 5 fNIRS studies related to PD, the activation of PFC in individuals with PD was higher than that in older adults during the walking task performed at a comfortable pace [17]. These findings suggest that walking is correlated with frontal brain activity in PD. The PFC is associated with cognitive regulation, working memory, and information synthesis, and all of these are necessary for sophisticated behavior [20]. Additionally, walking is influenced by posture control [21]. Posture control is primarily controlled by the sensory feedback system of the parietal lobe and further affected by other advanced cortical control functions such as sensorimotor integration, adaptation, and anticipation mechanisms [22]. The parietal lobe underlies behavioral control, intermodal integration of attention, and the processing of external incoming sensory signals [23]. Therefore, both the PFC and parietal lobes are involved in walking. However, how the PFC and parietal lobe are connected during walking is currently unknown. It is important to investigate the interaction patterns between the PFC and parietal lobe during walking.

Our study aimed to investigate the EC patterns between the PFC and parietal lobes during walking tasks performed by individuals with PD and healthy older individuals and to compare differences in the EC coupling strength of individuals with PD and healthy older individuals. We hypothesized that PD would show increased EC coupling strength compared to healthy older individuals because walking capabilities in PD may rely on compensation from cortical regions.

MATERIALS AND METHODS

Procedures

We evaluated the demographic characteristics and cognitive status of the two groups. The MMSE scale was used to evaluate overall cognitive function [27]. The Movement Disorders Society-United Parkinson’s Disease Rating Scale (MDS-UPDRS) Part III was used to evaluate the severity of motor symptoms [28]. The postural instability and gait difficulty (PIGD) score calculated based on the MDS-UPDRS was used to evaluate the PIGD severity [29]. The H-Y staging scale was used to determine the PD stage [26]. The most affected limb was determined based on the bilateral MDS-UPDRS sub scores.

Participants were required to walk at a normal and relaxed speed on a 10-m straight footpath. Each trial was started by standing for 30 s, followed by walking for 35 s, and stopped by standing on site for 10 s (Fig. 1). During this standing time, we asked participants to look straight ahead and not think of anything specific. The participants performed the same walking test three times. A 2-min rest period followed each walking test. Participants needed to complete the whole course. Individuals with PD were evaluated in the off-state because the off-state responds to the pathological state of PD and avoids the influence of drug status on the response.

OPEN IN VIEWER

Fig. 1

Experimental paradigm. Walking at a usual and relaxed speed on a 10-m footpath. Each trial started with standing for 30 s, followed by walking for 35 s, and stopped by standing on site for 10 s. The participants performed the same walking test three times.

fNIRS data recording

Brain signals were collected using a portable fNIRS system (Nirsmart; Danyang Huichuang Medical Equipment Co., Ltd., Danyang, China). The fNIRS system used near-infrared light at 730 and 850 nm to measure the optical intensities of oxygenated hemoglobin (HbO₂) and deoxygenated hemoglobin (HHb) at a sampling rate of 11 Hz; 26 fNIRS probes (14 sources and 12 detectors) were placed on the left prefrontal cortex (LPFC), right prefrontal cortex (RPFC), left parietal lobe (LPL), and right parietal lobe (RPL), resulting in 30 fNIRS channels (Fig. 2). The instrument was based on the principle of three-dimensional positioning and the international 10–20 system of electrode placement. The nasion, inion, and left and right preauricular points were anatomical reference points. The determination of 10–20 system regions of interest (ROI) was based on the standardization of different cap sizes from XS to L sizes to match different head circumferences. The source detector distance was 3 cm. The participants wore black hoods to reduce ambient light interference with light signals. Signal quality was checked by visual inspection.

OPEN IN VIEWER

Fig. 2

Schematic diagram of the experimental setup. Distribution of functional near-infrared spectroscopy (fNIRS) probes. Twenty-six fNIRS probes (14 sources and 12 detectors) were placed on the left prefrontal cortex (LPFC), right prefrontal cortex (RPFC), left parietal lobe (LPL), and right parietal lobe (RPL), resulting in a total of 30 fNIRS channels. Sources and detectors are shown as blue rectangles and green circles.

RESULTS

Table 1 summarizes the medical and population statistics characteristics. There was no significant difference in sex, age, height, and weight of individuals with PD and healthy controls (p > 0.05). There was no significant difference between the MMSE scores of individuals with PD and healthy controls (p > 0.05), indicating that the cognitive levels of individuals with PD and healthy controls in this study were similar. As expected, the MDS-UPDRS III scores showed statistical differences between the PD “off” and “on” states (p < 0.001).

Effective connectivity mCD pattern

Figure 3 depicts the interregional coupling direction during walking among the four evaluated cerebral regions of healthy controls and individuals with PD. No mCD was observed among the four cerebral regions during walking performed by the healthy controls. In the PD group, the main detected interregional coupling direction was from the LPL to the LPFC.

OPEN IN VIEWER

Fig. 3

Main coupling directions (mCDs) among brain regions in healthy controls (A) and individuals with Parkinson’s disease (B). LPFC, left prefrontal cortex; LPL, left parietal lobe; RFFC, right prefrontal cortex; RPL, right parietal lobe.

EC coupling strength

Figure 4 compares the healthy control and PD groups in terms of regional EC coupling strength. Compared with the healthy control group, the PD group had statistically significantly increased EC coupling strength in the connectivity types from LPL to LPFC (p = 0.008; effect size of Cohen d = –0.77), from LPL to RPFC (p = 0.010; effect size of Cohen d = –0.74), and from LPL to RPL (p = 0.018; effect size of Cohen d = –0.68).

OPEN IN VIEWER

Fig. 4

Comparison of regional effective connectivity (EC) between the healthy controls (HC) and Parkinson’s disease (PD) group. (A) Connectivity from the LPFC to the RPFC, LPL, and RPL. (B) Connectivity from the RPFC to the LPFC, LPL, and RPL. (C) Connectivity from the LPL to the LPFC, RPFC, and RPL. (D) Connectivity from the RPL to the LPFC, RPFC, and LPL. LPFC, left prefrontal cortex; LPL, left parietal lobe; RFFC, right prefrontal cortex; RPL, right parietal lobe. *p<0.05.

Figure 5 shows the EC matrices calculated from the fNIRS measurements of the healthy controls (A) and PD (B) group in the off state. The x- and y-axes of the square frames show the channel numbers within functional near-infrared spectroscopy (fNIRS), and the color at (x,y) indicates the value of the connectivity matrix (which represents the connectivity strength from channel x to channel y). Warmer colors indicate a higher connection strength. Compared to the healthy controls, the individuals with PD had a warmer color in the following channels: from the LPL to the LPFC; from the LPL to the RPFC; and from the LPL to the RPL.

OPEN IN VIEWER

Fig. 5

Effective connectivity (EC) matrices for healthy controls and PD group. The x-axes and y-axes of the square frames show the channel numbers within functional near-infrared spectroscopy (fNIRS), and the color at (x,y) indicates the value of the connectivity matrix (which represents the connectivity strength from channel x to channel y). LPFC, left prefrontal cortex; LPL, left parietal lobe; RFFC, right prefrontal cortex; RPL, right parietal lobe.

DISCUSSION

To our knowledge, this is the first study to investigate EC patterns in the cerebral cortex during walking tasks performed by individuals with PD and compare them with those of healthy controls. The mCD of EC reflects the driving relationships among different cerebral regions and the directions of these driving relationships [14]. This provides information regarding the functional hemodynamic responses in the brain. We observed that EC did not exhibit any mCD during walking tasks completed by healthy controls. This result supports the previous idea that, because of the long-term development of daily walking capabilities, this motor task has reached a level of automation that does not require complex cortex regulation in healthy individuals [38]. Previous studies of brain FC have shown that a lack of visible prefrontal activation is consistent with the automation of motor tasks [39]. Moreover, our findings support previous findings that, in healthy controls, motor control and parietal areas are less activated when exercise is automated [40].

In contrast, we observed that the mCD of individuals with PD occurred during walking from the LPL to the LPFC. Additionally, greater speed and greater stride length variabilities indicated that individuals with PD had decreased automaticity during walking tasks [4]. These results support previous observations that PD affects the subcortical pathway, resulting in dysfunction in automatic motion control and leading to the transformation of automatic gait control to cortical control [41, 42]. Moreover, compared with healthy controls, individuals with PD required higher activation of the cerebellum, premotor area, and parietal cortex to make automatic movements [41]. Additionally, the mCD from the LPL to the LPFC indirectly showed that the coupling direction from the LPFC to the LPL was weakened, indicating that the left prefrontal lobe is mainly regulated by the left parietal lobe during walking tasks in this population as well as demonstrating a decrease in connection transfer efficiency. The interpretation of mCD results may also be enhanced from LPL to LPFC, which is normal in the direction from LPFC to LPL. Since, there was no difference in the EC coupling strength from LPFC to LPL between the PD and healthy controls group.

In previous studies, the PFC was found to be more active during walking tasks performed by individuals with PD [43–45]. Moreover, according to a recent fNIRS study, PD inhibits LPFC activity during dual-task walking [46]. We also noted that the LPFC appears to be involved in working memory [47]. Although individuals with PD usually have impaired frontal function during frontal lobe-related tasks, research has shown that PD requires higher brain activity to perform automatic movements (including brain activity in the parietal cortex) [41]. Therefore, it seems that low-efficiency brain activity of individuals with PD is compensated for by increased premotor parietal activation during automatic motor execution. The current hypothesis also supports compensatory increases in premotor and parietal cortex activities in PD [48–50]. Our results suggest that individuals with PD are strongly dependent on the left prefrontal lobe to complete daily walking tasks, and that additional resources may be required from other brain regions as well (mainly regulated by the left parietal lobe).

Another possible explanation for this abnormal EC pattern in individuals with PD may be that individuals with PD have impaired sensorimotor integration. During the walking task administered during this study, the prefrontal cortex was involved in both execution and attention. Moreover, it is well-established that the parietal cortex is responsible for somatosensory perception [51], whereas the posterolateral parietal cortex is responsible for visual and spatial information [52, 53]. Traditionally, PD is considered a motor disorder, and sensory motor networks have a central role in the detection and processing of sensory input, as well as in the preparation and implementation of motor function [54]. Several studies have confirmed abnormal FC in the sensorimotor network (suggesting damaged sensorimotor integration) in PD [55, 56].

Another finding of our study was that EC coupling strength between the left parietal lobe and other brain regions was greater in individuals with PD than in healthy controls. The EC coupling strength from LPL to other brain regions was negatively correlated with gait speed, and positively correlated with speed variability in individuals with PD. Combined with the negative correlation between speed and speed variability in PD group, this suggested that the slower speed of participants with PD, the greater the speed variability, the more unstable the participant during walking, and the greater the impact of the LPL on other brain regions, as shown in Table 2. In the healthy controls group, we observed that EC types of RPFC as the major coupling generator were negatively correlated with speed. However, in the PD group, we found no correlation between the EC types of the PFC as the major coupling generator and behavioral performance. This might indicate that PFC activity is not as straight forward as previously suggested (i.e., more compensation, more activity).

A previous study using EEG measured the EC of eyes-open resting-state EEG signals in individuals with PD and healthy controls found obvious parietal-occipital separation; additionally, the information flow from the parietal and occipital regions to the motor cortex was statistically significantly reduced in individuals with PD [14]. This may be related to the results of individuals with PD who demonstrated decreased regional cerebral blood flow in the bilateral occipital lobe and posterior parietal cortex when compared with the control subjects [57]. Another study based on fMRI, the task was a dynamic balance task simulated on an fMRI-bed, found that the measurement of EC in PD during the process of dynamic balance indicated that individuals with PD had increased EC-associated movement in the parietal lobe and reduced connectivity from the brain stem to other subcortical areas [15]. Both the participants in the EEG and fMRI study were in the on state. Additionally, another research using PET investigated the brain correlates of motor timing task in PD, compared the EC between the off and on state, and found the analysis of EC demonstrated that excessive inhibitory pallidus outflow in the off state is associated to a lack of sufficient frontal activation and dependence on the cerebellum during the motor timing task [16]. In this study, participants with PD also need to lie down to complete motor timing tasks. However, because of the different methods (e.g., regarding imaging technology and task paradigms) implemented in previous studies, cannot meet the condition of definitive brain signal measurements in the walking state, since this process is a standing motion state process. To complete different tasks, the activated brain regions should be different, and the interaction patterns between brain regions will also be different. Moreover, in these studies, participants with PD were in the on state, which may not reflect the pathological effects of PD, because drugs can also affect cortical activity and task performance. This may limit our further exploration of the EC patterns during walking task based on previous studies. In our study, we found that EC of the left parietal lobe increased in individuals with PD. We explain this finding as a manifestation of compensatory hyperfunction in the left parietal lobe. Further research should be done on the correlation between cortex interaction and walking performance, including studying walking under different conditions and in different drug states, to provide a more thorough understanding and better explanation.

Additionally, an interesting finding of the current research was that we observed asymmetrical EC patterns on the left and right sides of the brain. This asymmetric distribution may have occurred because all recruits were right-hand-dominant, and the left hemisphere was the dominant hemisphere. In previous studies, the dominant hemisphere function of the sensory motor network in PD showed a tendency toward neurodegeneration. Moreover, the connection strength of the left brain network in PD has been found to decrease with aggravation of the disease [58]. Connection enhancement also has pathological significance. For instance, a previous study found that the auxiliary motor area and the right primary motor area in PD increased FC with regard to the motor cortex [59], which may be reflective of compensatory functional reorganization after long-term dopamine treatment [60]. Moreover, researchers conducted an fMRI study of handedness and found that the tendency toward handedness seemed to be associated with asymmetry in the frontal and parietal liaison pathways [61]; additionally, activation of this network showed a left bias in right-hand-dominant participants. In our study, the individuals with PD were relatively balanced in terms of the most severely affected side; however, all individuals with PD were right-hand-dominant, thus limiting the possibility of additional exploration of this topic. Therefore, we recommend that future studies should focus on this issue.

Since fNIRS can produce noisy results, it is important that the validity of the entropy measure is accounted for. In previous studies, Granger causality and dynamic causality models have been used to study effective connectivity in neuroscience [62, 63]. Compared with these methods, the characteristics of the transfer entropy method are nonlinear, and are independent of the established model and quantitative analysis [64]. The transfer entropy model has only recently been discovered, and there are still many unknowns. We hope that this study can provide some additional knowledge on the topic.

There were some limitations to our study. First, the fNIRS system can only detect signals in the cerebral cortex and cannot detect the deep structure of the brain, thus limiting our understanding of the putative mechanisms underlying our findings. Second, the spatial resolution of fNIRS is narrow, which restricts more accurate positioning of the brain region of interest; this resulted in a slightly broader partition of the cerebral cortex and a less detailed brain partition in the present study. We aim to refine this partition more comprehensively during our next research effort. Third, there were technical limitations of the filtering technology we used for the analysis. The technology we used was able to filter the noise with the bandpass Butterworth of 0.01–0.2 Hz; however, Mayer waves typically have a frequency of 0.1 Hz. Therefore, it is possible that physiological noise, such as signal contribution from increased blood flow in the scalp, may have influenced the results. We recommend that future studies use filtering technologies, such as accelerometers and short channels, that can appropriately filter out Mayer waves. Finally, we did not distinguish the process of gait initiation, termination, and turning during the walking tasks during the analysis. We may study the different walking processes in the future. Despite these limitations, our study is innovative and well-conducted, and it provides novel and critically important information.