EEG of Pre-movement Phase Aids Brain-Computer Interface (BCI) in Recognizing Movement Intentions

Background and Research Objectives

Brain-Computer Interface (BCI) is a technology that translates human intentions directly through neural signals to control devices, holding extensive application prospects [1]. BCI has the potential to revolutionize various fields such as daily life, entertainment, communication, rehabilitation, and education. However, current BCIs based on movement intentions face some challenges, particularly the less prominent EEG features during the pre-movement phase and susceptibility to attention influences, which hinder the performance enhancement of movement BCI.

Given this background, Yuxin Zhang, Mengfan Li, Haili Wang, Mingyu Zhang, and Guizhi Xu (corresponding author) from the School of Health Science and Biomedical Engineering at Hebei University of Technology, the State Key Laboratory for Reliability and Intelligence of Electrical Equipment, and the Tianjin Key Laboratory for Bioelectromagnetic Technology and Intelligent Health, conducted a study to investigate how to include the preparatory state during pre-movement coding and validate its feasibility in improving movement intention detection. Their research findings were recently published in Journal XX.

Research Methods and Experimental Procedure

The study designed two button tasks to induce participants to enter preparatory states for two movement intentions (left or right) under visual guidance. The experiment involved 14 participants (3 females and 11 males, right-hand dominant, aged between 23 and 28). The experimental procedure included two sub-experiments: spontaneous pre-movement and preparatory pre-movement.

Experimental Design and Data Acquisition

1. Spontaneous Pre-movement Experiment: After 3 seconds of relaxation, participants freely chose any moment within 10 seconds to press the keyboard button. Each press was followed by a 2-second rest period, with a total of 60 left-hand and 60 right-hand finger presses conducted.
2. Preparatory Pre-movement Experiment: After 3 seconds of relaxation, a left or right arrow appeared on the screen guiding the upcoming movement direction. After 1 second, the arrow switched to a progress bar, which lasted for 2 seconds until reaching 100%, upon which participants pressed the button followed by a 2-second rest period. A total of 120 trials were conducted (60 left-hand and 60 right-hand).

The experiments were conducted in a quiet environment, with EEG signals recorded using a 64-electrode portable wireless EEG amplifier. Various signal processing techniques (e.g., independent component analysis and common average reference) were employed to remove artifacts and noise.

Feature Extraction and Classification

The study extracted EEG data of low-frequency movement-related cortical potentials (MRCPs) and high-frequency event-related desynchronization (ERD) and used task-related spatial pattern (TR-CSP) and common spatial pattern (CSP) algorithms to merge different features for classification analysis. The study also compared the differences in time-domain, frequency-domain, and classification accuracy under different pre-movement conditions.

Main Research Results

MRCPs Features

The study found that preparatory pre-movement elicited lower amplitudes and earlier latencies in both contralateral and ipsilateral motor cortices. The contralateral dominance was notable, especially in the MRCPs waveforms induced by different tasks in low-frequency scenarios, showing significant differences in activation levels between the left and right brain regions.

ERD Features

In the frequency domain, ERD values induced by preparatory pre-movement were lower and quickly returned to baseline levels after button presses. Time-frequency curves in alpha and beta bands showed that preparatory pre-movement returned to baseline levels faster, demonstrating more apparent ERD features.

Classification Performance

By integrating features and classification methods, the study found that the classification accuracy under preparatory pre-movement conditions increased to 83.59%, significantly higher than the 78.92% under spontaneous pre-movement conditions (p<0.05). Additionally, the standard deviation reduced from 0.95 to 0.68, indicating that the proposed method had broader applicability and higher stability across different participants.

Conclusions and Significance

This study introduced a preparatory state into pre-movement coding and compared it with traditional spontaneous pre-movement, validating that the new coding paradigm significantly enhances neural representations of pre-movement. This method not only improves the detection performance of movement BCI but also expands the range of decodable movement intentions, holding significant application value.

Research Highlights

1. Innovative Coding Paradigm: Introducing a preparatory state significantly enhances pre-movement neural representations, providing new insights for improving BCI accuracy.
2. Cross-subject Stability: The paradigm demonstrated higher stability and applicability across multiple subjects, offering new solutions to BCI instability issues.
3. Application Value: The study showed that the new paradigm is suitable for rehabilitation and has the potential to expand the application range of movement BCIs.

This research demonstrates that incorporating a preparatory state in the pre-movement coding paradigm not only enhances classification accuracy and stability but also offers new possibilities for BCI applications, possessing broad scientific significance and practical application value.