The Promise of Deep Learning for BCIs: Classification of Motor Imagery EEG using Convolutional Neural Network

Abstract

Motor Imagery (MI) is a mental process by which an individual rehearses body movements without actually performing physical actions. Motor Imagery Brain-Computer Interfaces (MI-BCIs) are AI-driven systems that capture brain activity patterns associated with this mental process and convert them into commands for external devices. Traditionally, MI-BCIs operate on Machine Learning (ML) algorithms, which require extensive signal processing and feature engineering to extract changes in sensorimotor rhythms (SMR). However, in recent years, Deep Learning (DL) models have gained popularity for EEG classification as they provide a solution for automatic extraction of spatio-temporal features in the signals. In this study, EEG signals from 54 subjects who performed a MI task of left- or right-hand grasp was employed to compare the performance of two MI-BCI classifiers; a ML approach vs. a DL approach. In the ML approach, Common Spatial Patterns (CSP) was used for feature extraction and then Linear Discriminant Analysis (LDA) model was employed for binary classification of the MI task. In the DL approach, a Convolutional Neural Network (CNN) model was constructed on the raw EEG signals. The mean classification accuracies achieved by the CNN and CSP+LDA models were 69.42% and 52.56%, respectively. Further analysis showed that the DL approach improved the classification accuracy for all subjects within the range of 2.37 to 28.28% and that the improvement was significantly stronger for low performers. Our findings show promise for employment of DL models in future MI-BCI systems, particularly for BCI inefficient users who are unable to produce desired sensorimotor patterns for conventional ML approaches.

Functional Connectivity Analysis in Motor-Imagery Brain Computer Interfaces

“…Furthermore, classification scores of low performing users can be improved by incorporating new AI algorithms such as deep learning methods on raw EEG signals instead of the classical machine learning approach that relies on EEG feature extraction (e.g., Stieger et al, 2021;Tibrewal et al, 2021;Zhang et al, 2021). Deep learning models have the advantage of facilitating end-toend learning; they can exploit information from raw data on their own, which is not only computationally more effective but also captures brain activity patterns underlying MI beyond the defined ERD features (Tibrewal et al, 2021).…”

“…Furthermore, classification scores of low performing users can be improved by incorporating new AI algorithms such as deep learning methods on raw EEG signals instead of the classical machine learning approach that relies on EEG feature extraction (e.g., Stieger et al, 2021;Tibrewal et al, 2021;Zhang et al, 2021). Deep learning models have the advantage of facilitating end-toend learning; they can exploit information from raw data on their own, which is not only computationally more effective but also captures brain activity patterns underlying MI beyond the defined ERD features (Tibrewal et al, 2021). Particularly, in the light of connectivity research, deep learning provides a more holistic analysis of brain activity patterns during MI that extends beyond mu suppression in the sensorimotor area and this can result in a better discriminative power of BCI classifier for the inefficient users.…”

Improving Motor Imagery-Based Brain-Computer Interface Performance Based on Sensory Stimulation Training: An Approach Focused on Poorly Performing Users

“…BCI inefficiency is a significant problem that warrants research effort if these systems are to be useful in the future (Maskeliunas et al, 2016). Recent studies have tried to improve the classification performance of BCI inefficiency subjects using the deep learning method (i.e., convolutional neural network) because they cannot produce stronger contralateral ERD/ERS activity (Zhang et al, 2019b;Stieger et al, 2021;Tibrewal et al, 2021).…”

Artificial Intelligence in Motor Imagery-Based Bci Systems: A Narrative

“…Transferring source data from the old MI command makes this algorithm capable of performing efficiently for fewer training samples of the new MI command. However, this method is only suitable for discriminating between two classes [12]. "Discrete Wavelet Transform with Maximum Overlap" method was suggested by Abdul Wahab (2021) to categorize the theta and delta patterns of very low and high frequencies of brain activities during mediation, learning, sleeping or focus mode.…”

“…An electroencephalogram or Brain computer-based system measures specific features of an individual's brain signal, which correlates with their intention to affect control. [12] The system then translates such features into control signals to control external devices. Such technology is especially used in the medical field for neuro-muscular disorder patients, such as patients with "Amyotrophic Lateral Sclerosis (ALS), brainstem stroke, brain or spinal cord injury, cerebral palsy, muscular dystrophies, and multiple sclerosis".…”