Objective: This project aims at creating a tool for automatic EMG event- detection to support clinicians in this process.

Background: Neurophysiological evaluation of voluntary and involuntary movements in various movement disorders includes the analysis of cerebral activity by means of electroencephalography(EEG). A way to perform this analysis is by averaging EEG signal relative to a voluntary or involuntary movement recorded through electromyography(EMG). This technique, termed back-averaging, identifies brain patterns related to a particular motor activity and may elucidate its neural basis. Typically, these EMG events are manually labeled by expert clinicians in a time consuming process.

Methods: We developed a computer application that uses machine-learning methods to automatically detect EMG events and analyze their relation with cortical activity (i.e. EEG-EMG back-averaging). The application builds statistical models based on recordings already labeled by expert clinicians, and exploits them to speed up the event detection and analysis. We have explored the possibility of discriminating voluntary and involuntary movements using supervised machine learning methods. The performance of the system is being prospectively assessed by comparison with manual labeling of recordings involving both healthy subjects and patients with movement disorders.

Results: The application provides the visualization of the EEG/EMG recording – both in the temporal and spectral domain—, automatic detection of the EMG events and manual editing of such events. Automatic event detection showed high sensitivity with a low rate of false detections when compared to manually labeled events. Back-averaging of automatically detected voluntary movements in healthy subjects yielded the expected movement-related cortical potential (Bereitschaftspotential).

Conclusions: An automatic application has been developed for supporting the clinical evaluation of motor control and movement disorders. This tool provides automatic detection of EMG events based on statistical models. These models can be further refined by integrating samples labeled by an expert. Preliminary results on the discrimination of voluntary movements are encouraging, but further work is necessary to fully assess this approach for different movement disorders.

Annual Meeting of the Swiss Neurological Society, October 2015, in Berne Switzerland.

« Back to 2016 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/automatic-event-detection-for-the-neurophysiological-evaluation-of-voluntary-and-involuntary-movements/