A Novel Transfer Learning-Based Hybrid EEG-fNIRS Brain-Computer Interface for Intracerebral Hemorrhage...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Rehabilitation Medicine

Background:

• Motor imagery (MI) neurorehabilitation shows potential for intracerebral hemorrhage (ICH) recovery.
• Unimodal brain-computer interfaces (BCIs) struggle with cross-subject generalization due to neurophysiological heterogeneity.

Purpose of the Study:

• To develop and validate a multimodal EEG-fNIRS fusion framework for improved cross-subject generalization in MI-based neurorehabilitation for ICH.
• To introduce a Wasserstein metric-driven method for selecting optimal source domains to quantify inter-subject neural distribution divergence.

Main Methods:

• A multimodal framework fusing electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS).
• A Wasserstein metric-driven source domain selection method for quantifying inter-subject neural distribution divergence.
• Comparative neuroactivation analysis of 17 normal controls and 13 ICH patients during MI tasks using transfer learning.

Main Results:

• The transfer learning model achieved 74.87% mean classification accuracy on ICH patient data when trained with optimally selected normal templates.
• Cross-validation on public hybrid EEG-fNIRS datasets demonstrated generalizability, increasing baseline accuracy to 82.30% and 87.24%.

Conclusions:

• The proposed multimodal system synergistically combines EEG's temporal resolution and fNIRS's spatial specificity for ICH rehabilitation.
• This establishes the first clinically viable multimodal analytical protocol for ICH rehabilitation, advancing neurotechnology translation for personalized regimens.