EEGScaler: A Deep Learning Network to Scale EEG Electrode and Samples for Hand Motor Imagery Speed Decoding
None:
Motor Imagery (MI)-based Brain-Computer Interface (MI-BCI) systems induce neuroplasticity, promoting rehabilitation in stroke patients. Existing MI-BCI systems decode bilateral MI actions from Electroencephalogram (EEG) data to facilitate motor recovery. However, such systems offer limited degrees of freedom. Decoding kinematics information, such as movement speed can enhance control and provide a more natural interface with the environment. Decoding speed-related information from unilateral MI tasks is challenging due to the significant spatial overlap of neuronal sources and the inherently low spatial resolution of EEG. To address this, we propose EEGScaler, an end-to-end deep learning framework designed to decode slow v/s fast MI tasks by adaptively scaling EEG samples and electrodes with high discriminative value. EEGScaler leverages a Multi-Layer Perceptron (MLP) network to assign scale factors to both samples and electrodes. Spatiotemporal features are subsequently extracted using temporal and depth-wise convolution filters. The model is pre-trained on subject-independent data to learn filter weights, while subject-specific fine-tuning further optimizes the MLP-based scaling mechanism. The EEGScaler model performance is evaluated on 14 healthy subjects' data recorded while performing slow v/s fast unilateral MI tasks. The proposed model achieves an average cross-validated accuracy of 65. 98% for decoding fast v/s slow MI speed tasks, outperforming existing methods by approximately 6%. The subject-specific scaling of samples and electrodes using an end-to-end deep learning model for speed from unilateral MI tasks is novel. By effectively decoding movement speed, EEGScaler enhances the degree of freedom in MI-BCI systems, paving the way for more intuitive and efficient neurorehabilitation applications.Clinical Relevance- This advancement has the potential to improve motor rehabilitation strategies by enabling more precise and adaptive BCI-driven therapy tailored to individual recovery needs.
