Decoding Articulation Motor Imagery Using Early Connectivity Information in the Motor Cortex: A Functional...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Brain-computer interfaces (BCIs) offer communication pathways for individuals with motor impairments.
• Functional near-infrared spectroscopy (fNIRS) is a portable, non-invasive neuroimaging technique suitable for BCIs.
• Existing fNIRS BCIs often overlook neural functional connectivity, focusing primarily on activation patterns.

Purpose of the Study:

• To develop and evaluate a 4-class speech imagery BCI using fNIRS.
• To decode simplified articulation motor imagery (jaw and lip movements) of different vowels.
• To investigate the efficacy of functional connectivity features for improving speech BCI performance.

Main Methods:

• Utilized fNIRS to record brain activity during simplified vowel articulation imagery.
• Extracted synchronization information from the motor cortex as features.
• Implemented multiclass and binary classification analyses to decode vowel imagery.

Main Results:

• Achieved mean subject-dependent classification accuracies exceeding 40% in multiclass settings (4 vowels) within 0-2.5s and 0-10s time windows.
• Attained mean subject-dependent classification accuracies over 70% in binary settings (pairwise vowel comparisons) within the same time windows.
• Demonstrated that connectivity features effectively differentiate vowels even with reduced time windows (2.5s vs. 10s), with comparable decoding performance.

Conclusions:

• Functional connectivity features significantly enhance the decoding of vowel articulation imagery in fNIRS-based BCIs.
• Reduced time windows are feasible for effective decoding, suggesting optimization potential for real-time applications.
• Emphasizing articulation motor imagery from the motor cortex is crucial for advancing speech BCI technology.