Evaluation of Post-Stroke Impairment in Fine Tactile Sensation by Electroencephalography (EEG)-Based Machine Learning

Electroencephalography (EEG)-based measurements of fine tactile sensation produce large amounts data, with high costs for manual evaluation. In this study, an EEG-based machine-learning (ML) model support vector machine (SVM) was established to automatically evaluate post-stroke impairments in sensation. Stroke survivors (n = 12, stroke group) and unimpaired participants 15, control received stimulations cotton, nylon, wool fabrics the different upper limbs a participant dominant side control. The average maximal values relative spectral power (RSP) EEG were used as inputs SVM-ML model, which first optimized classification accuracies limb sides through hyperparameter selection (γ, C) radial basis function (RBF) kernel cross-validation during cotton stimulation. Model generalization investigated by comparing limbs. highest achieved (γ 21, C 23) RBF (76.8%) six-fold (75.4%), respectively, gamma band stimulation; these selected optimal parameters model. generalization, significant differences fabric stimulation shifted higher (beta/gamma) bands. generated results similar evaluation cortical responses stimulations; may aid automatic assessments sensations.