Practical electroencephalography (EEG) applications in stroke rehabilitation: Towards brain-computer interface (BCI) setup and motor function assessment

Electroencephalography (EEG) records electrical brain activity typically in a non-invasive manner. Recent literature has shown its potential in stroke rehabilitation, to actively engage stroke survivors in rehabilitation. In Chapter 3 of this thesis, the problems of EEG applications in stroke rehabilitation were firstly identified with a pilot study. Two main challenges were identified, hindering further application of EEG in stroke rehabilitation training. One of the challenges is that the BCI involved rehabilitation process is unsatisfying. Three objectives were derived from this challenge. Firstly, at the beginning of all EEG related stroke rehabilitation training, it is both time and effort consuming to go through data collection and model training for every rehabilitation task. Therefore, in Chapter 4 of the thesis, the possibility of using an EEG model from one type of motor imagery (e.g.: elbow extension and flexion) to classify EEG from other types of motor imagery activities (e.g.: open a drawer) was investigated. Secondly, a novel training method was introduced together with a rehabilitation platform in Chapter 5. The results suggested that the proposed methods in this thesis are feasible and potentially effective. Thirdly, the transition of the offline analysis method to an online control method is one of the major factors that affect BCI performance. However, research particularly focused on the method of filtering the prediction of an online classification is scarce. In Chapter 6, two methods of filtering online classification predictions were proposed and evaluated in a pseudo-online classification paradigm, with the EEG data collected from Chapter 5. The other challenge is related to motor function assessments in rehabilitation. Motor function is generally assessed with standard questionnaire-based assessments. In these assessments, the rater requires the ratee to perform pre-defined movements and gives scores based on the quality of the movements. Therefore, those motor function assessments have inevitable subjective influences on the functional scores. In Chapter 7 of the thesis, the author investigated the possibility of using EEG data to assess motor function. As a preliminary investigation, EEG-based motor function assessments were only investigated for upper-extremity among participants with stroke. The results suggested that EEG data can be used to assess motor function accurately.