Tripolar Concentric Ring Electrode Brain Computer Interfacing with Real and Imagined Movements

Author: Joshua Barclay

Barclay, Joshua, 2023 Tripolar Concentric Ring Electrode Brain Computer Interfacing with Real and Imagined Movements, Flinders University, College of Science and Engineering

Terms of Use: This electronic version is (or will be) made publicly available by Flinders University in accordance with its open access policy for student theses. Copyright in this thesis remains with the author. You may use this material for uses permitted under the Copyright Act 1968. If you are the owner of any included third party copyright material and/or you believe that any material has been made available without permission of the copyright owner please contact with the details.


Neuromuscular disorders such as multiple sclerosis or motor neuron disease lead to long-term degeneration of the efferent nervous system, resulting in progressive loss of motor function. Loss of motor function has been shown to severely impact independence and quality of life. As the underlying mechanisms triggering neuromuscular disorders are poorly understood, treatment focuses on improving patient independence and quality of life with assistive technologies. Brain computer interfaces (BCI) are a type of assistive technology that allow users to interact with exterior devices using brain activity alone, being particularly suited for patients with limited motor function. Many types of ‘paradigm’ can be used to evoke specific patterns in the brain, which can be used to control a brain computer interface. The motor imagery paradigm is one type, which requires the user to imagine a movement, triggering an event related desynchronisation (ERD) to occur within the relevant region of the sensorimotor cortex. Motor imagery paradigms are advantageous in that they allow intuitive control of a brain computer interface through self-modulation of their brain activity. Despite this, the current literature reports poor performance due to higher training requirements and reported BCI illiteracy. The proposed method was a modified motor imagery paradigm, which used real and imagined movements to train a classifier. Using EEG and EMG, event related desynchronisation was to be measured and recorded across movements, for use in training support vector machine and neural network classifiers. By using tripolar concentric ring electrodes (tCRE) as the sensory modality, it was hypothesised that this would reduce the presence of muscle artefacts, improving classifier training outcomes. Participants for the study were recruited from within the research group (n = 7). Participants were tasked with performing a series of movements, classified as either full extension, partial extension and imagined extension of the fingers. All participants demonstrated some level of event related desynchronisation using both emulated EEG and tCRE. From the channel demonstrating the greatest desynchronisation in each participant, a dataset was created for classifier training. A support vector machine was trained using leave-one-sample-out cross-validation, with a reported classification accuracy of (66.8%±3.71) and (65.6%±1.69) for emulated EEG and tCRE, respectively. Similarly, a neural network was trained using K-fold cross-validation, returning an emulated EEG accuracy of (51.7%±1.01) and tCRE accuracy of (52.7%±0.90). The results indicate that tCRE offers no additional benefit to classifier performance over emulated and ordinary EEG. Comparing with the literature, it was noted that studies utilising similar methods achieved higher classifier accuracy. It was speculated that this discrepancy was a result of the number of channels used for training the classifier. Support vector machine training was repeated, including all channels, with a reported accuracy of 88%, providing support for this speculation. Future studies should investigate the relationship between channel number and classifier performance further, particularly focusing on methods that maintain performance with a reduced channel setup.

Keywords: brain computer interface, event-related desynchronisation, tripolar concentric ring electrode

Subject: Medical Biotechnology thesis

Thesis type: Masters
Completed: 2023
School: College of Science and Engineering
Supervisor: Kenneth Pope