Language Mapping using tEEG and EEG Data with Convolutional Neural Networks.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Traditional electroencephalography (EEG) has limitations in signal quality and spatial resolution.
• Tripolar electroencephalography (tEEG) has demonstrated improvements in signal-to-noise ratio, spatial resolution, and high-frequency content compared to EEG.
• Accurate mapping of language dominance is crucial for neurosurgical planning, particularly for resective surgery.

Purpose of the Study:

• To compare the efficacy of tripolar electroencephalography (tEEG) and traditional electroencephalography (EEG) in mapping language lateralization to the left and right hemispheres.
• To analyze the performance of tEEG and EEG signals in both time and frequency domains using convolutional neural networks (CNNs).

Main Methods:

• Simultaneous acquisition of tEEG and EEG signals during language tasks.
• Application of convolutional neural networks (CNNs) for analyzing signal features.
• Comparison of hemispheric classification performance based on time-domain and frequency-domain features of both tEEG and EEG.

Main Results:

• Time-domain features of tEEG and EEG yielded comparable results for functional brain mapping.
• Frequency-domain features of tEEG and EEG showed significant differences.
• tEEG demonstrated equivalent left and right hemisphere classification performance in both time and frequency domains.
• EEG frequency-domain classification exhibited lower accuracy compared to tEEG.

Conclusions:

• Tripolar electroencephalography (tEEG) provides robust and accurate mapping of language dominance, outperforming traditional EEG in the frequency domain.
• The findings suggest tEEG is a promising non-invasive tool for quickly guiding clinicians on language dominance prior to resective surgery.
• tEEG's enhanced signal characteristics and consistent performance offer a valuable advancement for pre-surgical language mapping.