EEG-fTCD hybrid brain-computer interface using template matching and wavelet decomposition.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Brain-computer interfaces (BCIs) offer novel communication pathways for individuals with severe motor impairments.
• Hybrid BCIs integrating multiple neuroimaging modalities can potentially improve performance over single-modality systems.
• Previous work established a foundational hybrid BCI using electroencephalography (EEG) and functional transcranial Doppler (fTCD).

Purpose of the Study:

• To enhance the accuracy and information transfer rate (ITR) of a hybrid EEG-fTCD BCI.
• To refine feature extraction and implement a Bayesian fusion approach for improved signal integration.
• To evaluate the BCI's performance in discriminating between cognitive tasks: mental rotation (MR) and word generation (WG).

Main Methods:

• Extended feature extraction using template matching for EEG and multi-scale wavelet decomposition for fTCD.
• Applied Wilcoxon signed rank test for significant feature selection.
• Utilized Support Vector Machines (SVM) for classification and proposed a Bayesian fusion of EEG and fTCD data.
• Evaluated performance on MR, WG, and baseline tasks.

Main Results:

• Achieved high classification accuracy: 98.11% for WG vs. MR, 86.27% for MR vs. baseline, and 85.29% for WG vs. baseline.
• Reported significant improvements in ITR, with MR/WG vs. baseline ITRs doubling and WG vs. MR ITRs quadrupling compared to previous studies.
• The hybrid EEG-fTCD BCI demonstrated superior performance compared to existing EEG-fNIRS BCIs.

Conclusions:

• The developed analysis techniques substantially improved hybrid BCI performance.
• The Bayesian fusion approach effectively integrates EEG and fTCD data, leading to enhanced accuracy and ITR.
• This advanced hybrid BCI represents a significant step forward in neurotechnology, outperforming previous systems.