Time Synchronization Between Parietal–Frontocentral Connectivity with MRCP and Gait in Post-Stroke Bipedal Tasks

Time Synchronization of Motor-Related Cortical Potentials and Parieto-Frontocentral Connectivity in Bilateral Tasks of Stroke Patients

Background

In stroke rehabilitation research, functional connectivity (FC), motor-related cortical potentials (MRCP), and gait activities are common metrics related to rehabilitation outcomes. Although these have been studied individually, their interrelationships, especially in the context of bilateral differentiation, have not been deeply explored. The significant variation in rehabilitation outcomes among stroke patients indicates that understanding the relationships among these metrics could reveal new strategies and therapies for rehabilitation.

Paper Source

This paper was written by Chun-Ren Phang, Kai-Hsiang Su, Yuan-Yang Cheng, Chia-Hsin Chen, and Li-Wei Ko, from institutions such as National Yang Ming Chiao Tung University and Kaohsiung Medical University. It was published in the “Journal of Neuroengineering and Rehabilitation” in 2024.

Research Process

Experimental Design

In this study, 10 participants were equipped with EEG devices and inertial measurement units (IMUs) while performing lower limb motor preparation (MP) and motor execution (ME) tasks. Data was collected to extract MRCP, functional connectivity (FCs), and bilateral differentiation, and to calculate knee angle changes during the ME phase from gait data. Pairwise Pearson correlation analysis was used to determine functional connectivity, and whole-brain connectivity data was fed to a support vector machine (SVM) for bilateral classification.

Data Collection and Processing

The experiment used a wireless EEG device called St. EEGTM Vega, a 32-channel system produced by ARTISE Biomedical Co., Ltd. Electrodes were placed according to the International ¹⁰⁄₂₀ system, and data were collected using Cynus data acquisition software at a sampling frequency of 500 Hz. The average impedance of EEG electrodes was maintained below 100 kΩ. IMUs were attached to the patient’s waist, thighs, and lower legs to capture knee angle changes during the ME phase.

Data Analysis

• MRCP (Motor-Related Cortical Potentials): EEG signals were processed using a 0.1-5 Hz bandpass FIR filter, and averaged across trials. The desynchronization trough values were compared during movements of the hemiplegic and non-hemiplegic legs.
• Functional Connectivity and PFCC (Parieto-Frontocentral Connectivity): EEG signals were filtered using an 8-50 Hz bandpass FIR filter, and functional connectivity was estimated using pairwise Pearson correlation coefficients. Time-varying connectivity was calculated within different windows, specifically analyzing connectivity at P3-FC4 and P3-C4 regions related to motor performance.
• SVM Machine Learning: A linear SVM was used for classification tasks, with the functional connectivity matrix vectorized and input into the SVM. Ten-fold cross-validation was used to report the average accuracy for each participant. The performance of static and time-varying connectivity in bilateral classification was compared.
• Knee Angles: IMU data were used to extract and average angle changes across all trials, synchronized with the four features extracted from EEG data (MRCP, PFCCs, classification accuracy, and knee angle changes).

Research Results

Knee Angle Changes and MRCP

Typical MRCP waveforms were observed for all participants, marking a rapid decrease in EEG amplitude at the onset of movement, reaching a trough when the knee angle was at its maximum, and rebounding as the gait returned to its initial position. The hemiplegic foot showed greater desynchronization, approximately 0.25 μV higher compared to the non-hemiplegic foot during movement.

MRCP and PFCCs

Time synchronization of MRCP and PFCCs revealed that during the ME phase, a decrease in MRCP amplitude was accompanied by an increase in PFCCs strength. During the MP phase, PFCCs strength decreased. Both phases showed sensitive responses of PFCCs to motor tasks, with significantly increased PFCCs connectivity during the movement of the hemiplegic foot.

Bilateral Classification Accuracy

The time-varying classification accuracy indicated a significant increase in bilateral classification accuracy following the ME cue. All participants achieved classification accuracy above the random guess threshold before the onset of lower limb movement, demonstrating that connectivity features could distinguish between left and right foot motor preparation activities. Further analysis revealed that classification accuracy in the pre-movement phase was higher than during the MP phase, averaging 75.1%.

Discussion and Conclusion

These results demonstrate a negative correlation between changes in knee angle, MRCP desynchronization, and PFCCs enhancement. MRCP was more prominent during the ME phase, while PFCCs were more dependent on the MP phase. This finding provides better neurophysiological insights and reveals mechanisms by which PFCCs and MRCP complement each other during motor task execution.

The study further indicates that using PFCCs can monitor and evaluate the rehabilitation progress of the hemiplegic foot post-stroke, while the high classification accuracy in the premovement phase suggests that these connectivity features can be used for rapid and effective classification tasks. These findings lay the foundation for developing brain-exoskeleton interfaces that can help stroke patients control rehabilitation exoskeletons while monitoring central nervous system recovery.

Future research directions include adopting multi-domain EEG features and integrating machine learning models to improve bilateral classification performance, and applying these findings to develop brain-exoskeleton interface systems capable of real-time noise detection and cancellation, thereby enhancing rehabilitation outcomes and comfort.