Single-trial classification of motor imagery differing in task complexity: a functional near-infrared spectroscopy...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Rehabilitation Technology

Background:

• Brain-computer interfaces (BCIs) offer potential for neurorehabilitation by monitoring brain signals via non-invasive methods like functional near-infrared spectroscopy (fNIRS).
• Single-trial classification of these brain signals is crucial for effective BCI operation.

Purpose of the Study:

• To evaluate offline single-trial classification of brain signals using a novel wireless fNIRS instrument.
• To discriminate between simple and complex motor imagery (MI) tasks based on fNIRS-derived brain signals.

Main Methods:

• 12 subjects performed simple and complex MI tasks (right thumb vs. all fingers sequential tapping).
• fNIRS data were recorded over secondary motor areas.
• Fisher's linear discriminant analysis (FLDA) was used with cross-validation to select optimal features (channels, time intervals, Δ[O2Hb] signal features).

Main Results:

• An average classification accuracy of 81% was achieved in discriminating between simple and complex MI tasks.
• The classification utilized a combination of specific channels, time intervals (5-15s post-stimulus), and Δ[O2Hb] signal features (mean amplitude, variance, skewness, kurtosis).
• Results showed significant discrimination (p ≤ 0.001) over secondary motor areas.

Conclusions:

• The findings show promising classification accuracies for differentiating MI task complexities using fNIRS.
• Subject-to-subject variability in classification accuracy was noted, requiring further investigation.
• The study's results and limitations are discussed concerning future BCI development and neurorehabilitation applications.