Abstract
Electroencephalogram (EEG) is a neuroimaging technique that records the electrical activity of the brain non-invasively by measuring the voltage on the scalp. A wide range of studies on EEG has looked into oscillatory rhythms, connectivity and recruitment pattern in different tasks or events. Yet, transferring the knowledge from these multi-channel EEG studies to real-life applications remains a challenge, not only due to the time-consuming setup, but also because of the variance observed within the tasks and between the subjects. This study investigates whether a single EEG channel with information transfer to and/or from other channels can be identified for optimal performance whilst keeping the number of channels involved at minimum. It also examines the possible use of a wireless single-channel EEG headset for applications.Non-parametric Transfer Entropy (TE) was adopted in studying information transfer and connectivity across the neural system. Although TE is widely used in physics, there remains a lack of simulation or validation studies to investigate whether TE is applicable to EEG data. Therefore, this study first evaluated, through intensive simulations, TE with respect to different coupling strengths within the neural mass model. A comparative study on Granger Causality and Transfer Entropy was conducted in various conditions, such as different coupling strengths, frequencies and topologies. The results indicate that transfer entropy is more sensitive than Granger Causality in the case of weak coupling. This suggests that TE is more useful to weakly coupled electrophysiological signals, such as EEG.
The performance of TE in the study of information transfer from the Prefrontal Cortex, an area relatively easier to access for the EEG setup, was then validated. The study first examined TE in motor task, since Electromyography (EMG) signal could provide clear muscle activation during movement as a task performance indicator. The information transferred in and out at Fp1 signified a critical channel in motor task. The physical activities could be clearly reflected by EMG during hand movement for comparison. The results of this study demonstrate a direct information transfer from EEG to EMG during grasping. Since EEG and EMG have different ranges of oscillations, the signals were further decomposed into different frequency ranges via wavelet transformation to examine cross-frequency information transfer. As such, information transfer from specific EEG bands to the high frequency component of EMG changed instantaneously at the onset and end of the movement.
Thus far, multi-channel EEG analysis has provided some insight into the brain functions, but it is time-consuming to set up with limited use outside the laboratory. To investigate the feasibility of bringing the EEG technology to real-life applications, prefrontal EEG signals collected from a wireless single-channel device were analysed to evaluate the dynamics of brain activity. Following from the motor task, the mental workload in various cognitive tasks was examined. Short-time EEG analysis reveals the theta activity increase as the common feature of mental workload across different tasks. Meanwhile, the real-time mental workload could be classified from EEG features with 65% − 75% accuracy across subjects. This result suggests the possible use of mobile EEG device to facilitate the evaluation of mental effort in specific tasks.
| Date of Award | 13 Jan 2017 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Ho Man CHAN (Supervisor) |
Cite this
- Standard