Related Experiment Video
Updated: Oct 15, 2025
06:34
A Single-Channel and Non-Invasive Wearable Brain-Computer Interface for Industry and Healthcare
Published on: July 7, 2023
2.6K
Investigating Data Cleaning Methods to Improve Performance of Brain-Computer Interfaces Based on
Shengjie Liu1, Guangye Li1, Shize Jiang2
1State Key Laboratory of Mechanical Systems and Vibrations, Institute of Robotics, Shanghai Jiao Tong University, Shanghai, China.
Frontiers in Neuroscience
|October 25, 2021
Summary
Data cleaning methods significantly impact brain-computer interface (BCI) performance using stereo-electroencephalography (SEEG) signals. The Laplacian reference method demonstrated superior gesture decoding accuracy by enhancing low-frequency signal distinguishability.
Area of Science:
- Neuroscience
- Biomedical Engineering
- Signal Processing
Background:
- Stereo-electroencephalography (SEEG) records brain activity via depth electrodes, offering potential for brain-computer interfaces (BCIs).
- Effective noise reduction through data cleaning is crucial for SEEG signal analysis and BCI performance.
Purpose of the Study:
- To investigate the impact of various data cleaning methods on SEEG-based BCI decoding performance.
- To identify the underlying reasons for performance differences across cleaning techniques.
Main Methods:
- Applied five distinct data cleaning methods (common average, gray-white matter, electrode shaft, bipolar, Laplacian references) to SEEG data.
- Evaluated the effect of each method on gesture decoding accuracy.
- Analyzed signal changes in spatial, spectral, and temporal domains.
Main Results:
- All tested data cleaning methods improved gesture decoding accuracy compared to no cleaning.
- The Laplacian reference method yielded the highest decoding performance.
- Superior performance of the Laplacian reference correlated with enhanced signal distinguishability in the low-frequency band.
Conclusions:
- Proper data cleaning is essential for optimizing SEEG-based BCI systems.
- The Laplacian reference is a highly effective method for SEEG data preprocessing in BCI applications.
- Understanding signal domain changes aids in selecting optimal cleaning strategies.
