Brain-Computer Interface Channel-Selection Strategy Based on Analysis of Event-Related Desynchronization Topography in Stroke Patients

Chong Li, Tianyu Jia, Quan Xu, Linhong Ji, Yu PanJournal of Healthcare Engineering 2019 · DOI

TL;DR

This paper proposes a novel channel-selection strategy for Brain-Computer Interfaces (BCIs) in stroke rehabilitation. By analyzing the unique Event-Related Desynchronization (ERD) patterns in stroke patients, they demonstrate that an individualized approach significantly improves motor intention classification accuracy compared to traditional methods. This is crucial for effectively stimulating neural pathways and enhancing motor recovery in individuals with brain damage.

What Problem Does This Paper Address?

Stroke survivors often experience motor deficits, and technology-assisted rehabilitation, particularly using EEG-based BCIs, holds great promise. BCIs work by creating a closed loop between motor intention and actual movement, aiming to stimulate and retrain neural pathways. However, stroke-induced brain damage can cause motor intention signals to shift to different brain regions, a phenomenon known as neural reorganization. Traditional BCI channel selection often relies on the contralateral sensorimotor cortex (SM1), which may not be the primary area for motor intention expression in stroke patients with significant brain damage. This study addresses the challenge of identifying the most relevant EEG channels for accurate motor intention detection in stroke patients, considering their individualized patterns of brain activity and potential neural reorganization.

Methods

The researchers recruited 15 stroke patients and employed a cue-based experimental paradigm. Participants were instructed to attempt opening their paretic or unaffected hand in response to visual cues. Electroencephalogram (EEG) data was recorded using 64 electrodes.

The core of their methodological contribution lies in the channel-selection strategy. They compared three approaches:

  • SM1-4 and SM1-5: These are traditional strategies that select channels from the contralateral sensorimotor cortex (SM1), based on established practices for healthy individuals.
  • ERD Selection: This is their novel, individualized strategy. It involves analyzing the Event-Related Desynchronization (ERD) topography during motor tasks. ERD, a decrease in EEG power, is indicative of neural activity. This strategy selects the 4 channels exhibiting the lowest ERD ratios, hypothesizing that these channels best capture the patient's unique motor intention signal, even if it's in non-traditional brain regions.

After channel selection, EEG data was preprocessed (filtering, artifact removal) and subjected to time-frequency analysis (Morlet Wavelet) to calculate ERD. Common Spatial Pattern (CSP) was used for feature extraction, and a Support Vector Machine (SVM) classifier was employed for offline classification accuracy assessment using 5-fold cross-validation.

Datasets

The study utilized EEG data collected from 15 stroke patients during a motor attempt task. No new benchmark datasets were introduced; rather, the study focused on analyzing existing EEG data from a specific patient population.

Key Insights and Ideas

  • Individualized ERD Topography: Stroke patients exhibit highly personalized patterns of ERD during motor attempts, deviating from the typical SM1 activation seen in healthy individuals. This highlights the significant impact of brain damage and subsequent neural reorganization.
  • ERD as a Biomarker for Channel Selection: The ERD topography can effectively pinpoint the brain regions actively involved in motor intention, even if they are not the traditional SM1 areas. This provides a data-driven approach to selecting optimal channels.
  • Limitations of Traditional Channel Selection: Relying solely on the contralateral SM1 for channel selection can be suboptimal for stroke patients, especially those with extensive motor cortex damage, leading to lower classification accuracy.
  • Compensatory Brain Regions Matter: For patients with severe motor cortex damage, compensatory brain regions play a crucial role in motor control. The ERD selection strategy can identify these regions, enabling more accurate BCI control.

Key Results

  • Improved Classification Accuracy: The ERD selection strategy consistently outperformed traditional SM1-based channel selection methods. For the "ERD-proliferation group," ERD selection achieved an average accuracy of 79.38±10.28%, significantly higher than SM1-4 (70.00±9.35%) and SM1-5 (71.25±8.54%).
  • Statistical Significance: One-way ANOVA confirmed a statistically significant difference in classification accuracy between the channel-selection strategies (p < 0.0005). Post-hoc tests showed that ERD selection was significantly better than both SM1-4 and SM1-5.
  • Patient Group Variations: The performance gains varied across different patient groups based on their ERD patterns. The ERD-proliferation group showed the most substantial improvement with ERD selection.
  • ERD-Blind Patients: Some patients (S4, S11, S14) were identified as "ERD-blind," showing no significant ERD during motor tasks, which the authors excluded from further analysis.

Limitations and Future Work

  • Heterogeneous Patient Group: The study included patients with diverse lesion types, sites, and severity (indicated by FMA scores), making it challenging to establish definitive correlations between specific clinical factors and ERD patterns.
  • No Direct Correlation with FMA: The authors acknowledge the difficulty in finding a direct quantitative correlation between ERD topography and Fugl-Meyer Assessment (FMA) scores due to the heterogeneity of the patient group.
  • Future Research Directions:
    • Investigate larger cohorts with more homogeneous patient groups (e.g., similar lesion types) to establish stronger correlations.
  • Validate the actual rehabilitation effects of BCIs using the proposed individualized channel-selection strategy in clinical trials.
  • Explore the dynamic updating of channel selection as neural reorganization evolves during rehabilitation.

Summary updated

Abstract

In the last decade, technology-assisted stroke rehabilitation has been the focus of research. Electroencephalogram- (EEG-) based brain-computer interface (BCI) has a great potential for motor rehabilitation in stroke patients since the closed loop between motor intention and the actual movement established by BCI can stimulate the neural pathways of motor control. Due to the deficits in the brain, motor intention expression may shift to other brain regions during and even after neural reorganization. The objective of this paper was to study the event-related desynchronization (ERD) topography during motor attempt tasks of the paretic hand in stroke patients and compare the classification performance using different channel-selection strategies in EEG-based BCI. Fifteen stroke patients were recruited in this study. A cue-based experimental paradigm was applied in the experiment, in which each patient was required to open the palm of the paretic or the unaffected hand. EEG was recorded and analyzed to measure the motor intention and indicate the activated brain regions. Support vector machine (SVM) combined with common spatial pattern (CSP) algorithm was used to calculate the offline classification accuracy between the motor attempt of the paretic hand and the resting state applying different channel-selection strategies. Results showed individualized ERD topography during the motor attempt of the paretic hand due to the deficits caused by stroke. Statistical analysis showed a significant increase in the classification accuracy by analyzing the channels showing ERD than analyzing the channels from the contralateral sensorimotor cortex (SM1). The results indicated that for stroke patients whose affected motor cortex is extensively damaged, the compensated brain regions should be considered for implementing EEG-based BCI for motor rehabilitation as the closed loop between the altered activated brain regions and the paretic hand can be stimulated more accurately using the individualized channel-selection strategy.
BibTeX
@article{li2019brain,
  author = {Chong Li and Tianyu Jia and Quan Xu and Linhong Ji and Yu Pan},
  title = {{Brain-Computer Interface Channel-Selection Strategy Based on Analysis of Event-Related Desynchronization Topography in Stroke Patients}},
  journal = {Journal of Healthcare Engineering},
  year = {2019},
  doi = {10.1155/2019/3817124},
}
Figure