GRADUATE SCHOOL OF SCIENCES & ENGINEERING
BIOMEDICAL SCIENCES AND ENGINEERING
PhD THESIS DEFENSE BY GIZEM YILMAZ
Title: Contamination of the EEG by Postural Activity of Temporalis Muscle: Single Motor Unit Approach to Evaluate the Extent of Muscle Interference
Speaker: Gizem Yılmaz
Time: January 18, 2019, 10.30
Place: ENG 208
Rumeli Feneri Yolu
Thesis Committee Members:
Prof. Dr. Kemal S. Türker (Advisor, Koç University)
Prof. Dr. Pekcan Ungan (Koç University)
Prof. Dr. Tamer Demiralp (İstanbul University)
Prof. Dr. Yasemin Gürsoy Özdemir (Koç University)
Doç. Dr. Barış Isak (Marmara University)
From the diagnosis of neurological disorders in clinics to the state of the art brain computer interface systems, the electroencephalography (EEG) has constituted an important tool in monitoring brain’s ongoing activitiy. The scalp EEG is recorded with electrodes from few to hundreds in number (depending on the aim and the application) that are placed on the scalp to capture and record the electrical activity of neural sources. However, head is also covered with facial, masticatory and neck muscles whose activities can be mixed up with the neural signal. This signal contamination problem is eccentuated when the frequency overlap between the EEG and the electromyography (EMG) is considered. The myogenic contamination of the EEG has been well documented for the voluntary activation of muscles and via surface EMG recordings. However, first, surface EMG is not reliable alone to detect the source of myogenic activity because of the cross-talk problem. Second, facial, masticatory and neck muscles exhibit unintentional, involuntary activities which may have postural role.
Following the two drawbacks listed, here in this thesis we aimed to evaluate the extent of myogenic contamination during recordings at rest. It was shown before that single motor unit activity interferes with the EEG signal. Here, we first wanted to evidence the temporalis activity at rest is due to postural activity in maintaining mandibular posture. Secondly, we wanted to compose a simple, artefact like EEG signal from the linear summation of surface representations of temporalis single motor units. We further compared the EEG-like signal and the original EEG in terms spectral characteristics. We demonstrated the widespread interference from temporalis SMU activity over the EEG electrodes even at rest and the surface EMG of temporalis activity was contained in the scalp EEG signal. In the last step, we applied an artefact removal algorithm to clean the EEG signal from artefactual components such as eyeblinks and muscle activities. Our aim was to evaluate the SMU interference before and after artefact pruning to investigate if artefact removal can detect and remove the SMU level interference. We observed that the application of an Independent Component Analysis (ICA) based artefact removal algorithm on the EEG signal decreased the amplitude of interference but could not completeley rule it out.
Our findings so far illustrated that the SMU level interference from the tonic or postural activity of temporalis is quite prominent to ignore, contaminating EEG bands particularly in beta and gamma frequencies. Hence, the interference can be persistent after ICA processing. Considering that temporalis is only one of the muscles among many surrounding the head, the EEG-EMG relationship should be reconsidered cautiously. New approaches and removal algorithms with sensitivity to SMU level activity are needed for both clinicians and engineers in the EEG field.