A penalized time-frequency band feature selection and classification procedure for improved motor intention decoding...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Motor imagery brain-computer interfaces (MI-BCIs) leverage electroencephalography (EEG) for neurological patient assistance and rehabilitation.
• Current MI-BCI models often rely on the Common Spatial Patterns (CSP) algorithm, which is sensitive to predefined frequency bands and time segments.
• This sensitivity limits the adaptability and reliability of subject-specific decoding of motor intentions.

Purpose of the Study:

• To develop a novel, efficient decoding algorithm for EEG-based MI-BCIs that overcomes the limitations of traditional CSP methods.
• To enhance the accuracy and robustness of motor intention detection by analyzing multiple time-frequency bands.
• To enable more reliable BCI applications for neurological rehabilitation and assistance.

Main Methods:

• A new decoding algorithm decomposes raw multichannel EEG data into multiple temporal and frequency bands.
• Features are extracted within each band using CSP, followed by simultaneous feature selection and classification via elastic-net regression.
• The method was validated on public and self-acquired EEG datasets using single and multiple temporal window configurations.

Main Results:

• The proposed multiple time-frequency band method achieved accuracy improvements of up to [Formula: see text] (average 84.8%) compared to state-of-the-art methods.
• The approach demonstrated reduced classification variability, indicating enhanced robustness to intra-subject EEG fluctuations.
• Automatic selection of subject-specific spatio-temporal-spectral features was achieved, particularly for detecting motor imagery against rest conditions.

Conclusions:

• The novel method significantly improves motor intention detection accuracy and robustness in EEG-based MI-BCIs.
• This approach offers a more adaptable and reliable subject-specific model for decoding motor intentions.
• The technique advances the development and application of EEG-based MI-BCIs for neurorehabilitation and assistive technologies.