Early Detection Via EEG Signals of Self-Initiated Reaching and Grasping Movements Performed With The Subject's Selected Upper Extremity

• Published in: IEEE access Vol. 13; p. 1
• Main Authors: Sanchez-Bautista, Jorge A., Antelis, Javier M., Mendoza-Montoya, Omar
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.01.2025; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Accuracy; Brain; Classification; Discrete wavelet transforms; EEG signals; Electroencephalography; Frequencies; Grasping; Human motion; Limbs; Long short term memory; Motion perception; Motors; movement intention; Protocols; reach and grasp; self-paced movement; Support vector machines; upper limb; Visual stimuli; Visualization
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2025.3553554

Recognizing movement intention through electroencephalography (EEG) signals in decision-making tasks remains a challenge for developing systems that translate brain activity into commands for assistive applications. By understanding and incorporating this information, these systems could come closer to replicating human movement's natural and voluntary behavior. This research analyzed brain activity before self-initiated reach-and-grasp movements performed with a participant-selected upper limb to identify relevant information regarding movement onset and configuration. A classification analysis was then conducted to assess the performance of a prospective assistive system in detecting movement onset, identifying the limb used, and distinguishing the type of movement executed. EEG signals were recorded from 20 healthy participants aged 18-45 in an environment insulated from external noise and visual stimuli to minimize evoked potentials. When evaluating the upper limb used, the analyses indicated relevant information and significant differences in signals acquired from different electrodes, especially in the Alpha and Beta frequency bands. However, these differences diminished when comparing vertical and lateral movements. For movement onset classification, an average accuracy of 82% was achieved, while the limb classification reached 50% accuracy in a three-class scenario. For the classification of movement type, the average accuracy was 45%. These results suggest that detecting movement intention is feasible and decoding movement intentions could contribute to developing more natural and intuitive device-assisted neurorehabilitation therapies.