Improving Movement-Related Cortical Potential Detection at the EEG Source Domain

Abstract

Decoding movement-related cortical potential (MRCP) plays an important role in the Brain-computer interface (BCI) system. MRCP is not easy to be detected on sensors due to the volume conduction effect. This work combines the scout EEG source imaging (ESI) and Locality Preserving Projection (LPP) followed by a linear discriminant analysis (LDA) classifier to detect MRCP of motor imagery and execution. Seven healthy subjects participated in this study and performed cue-based ballistic dorsiflexion. Our results showed that the source domain-based method achieved a significantly higher true positive rate (TPR) than that obtained from the sensor domain in both motor imagery (MI) (76.65±4.26% vs. 70.3±5.4%) and motor execution (ME) (81.66±2.55% vs. 74.49±6.48%) tasks. The false-positive rate (FPR) calculated in the source or sensor space for MI and ME was (24.46±5.07% vs. 28.59±5.17%) and (22.52±3.35% vs. 25.99±7.37%), respectively. Therefore, we demonstrated that EEG signals obtained from the source domain could improve the MRCP detection rather than those in the sensor domain.