Home » Publication » 23617

Dettaglio pubblicazione

2021, Nature Conferences-Technologies for Neuroengineering, Pages -

Towards A Novel Hybrid Brain-Computer Interface for Motor Rehabilitation: Study on Cortico-Muscular Coherence Patterns for Movement Classification (04b Atto di convegno in volume)

De Seta Valeria, Colamarino Emma, Pichiorri Floriana, Masciullo Marcella, Cincotti Febo, Mattia Donatella, Toppi Jlenia

Hybrid Brain-Computer Interfaces (hBCIs) for upper limb motor rehabilitation after stroke should pursue the reinforcement of “more normal” brain and muscular activity. We propose cortico-muscular coherence (CMC) as control feature for a novel rehabilitative hBCI. CMC is a measure of synchronization between central and peripheral activation and we tested its ability in discriminate different motor tasks with the ultimate aim of addressing the design of a hBCI able to train both brain and peripheral activity. Electroencephalographic (EEG) signals from 31 electrodes placed over the sensorimotor area and surface electromyography (EMG) from 5 muscles per side were collected in 20 healthy participants performing finger extension (Ext) and grasping (Grasp) with both dominant and non-dominant hand (60 trials). Each trial consists of 3s of rest and 4s of motor execution. Data were pre-processed and segmented in epochs, selecting windows of 2s each for rest and task condition. Single-trial CMC was estimated for each EEG-EMG pair in the frequency bands of interest: alpha (8—12 Hz), beta (13—30 Hz), gamma (31—60 Hz) and high frequency (61—100 Hz). For each movement and frequency band, CMC values were extracted at the characteristic frequency in both task and rest condition and used as features to classify each task vs rest, and Ext vs Grasp in each limb, by means of a single-subject 10-iteration cross-validation. CMC values across frequency bands were considered both together (full-spectrum classifier) and separately (band-specific classifiers) in the feature space of different classification models. We demonstrated that CMC features allowed for classification of both movements vs rest with better performance (Area Under the receiver operating characteristic Curve, AUC) for the Ext movement (AUC 0.96) with respect to Grasp (AUC 0.88). Classification of Ext vs Grasp showed higher discriminability in beta and gamma frequency bands (AUC 0.96). Our preliminary findings indicated that CMC could provide a comprehensive framework of the physiological patterns (cortical and muscular) during simple hand movements to eventually be employed in a hBCI system for post-stroke rehabilitation.
© Università degli Studi di Roma "La Sapienza" - Piazzale Aldo Moro 5, 00185 Roma