In recent decades, transcranial magnetic stimulation (TMS) has emerged as an effective treatment for depression, as well as other psychiatric conditions. However, similar to antidepressants, only approximately 50% of patients treated with TMS respond to the treatment. TMS treatments are provided to each patient in the clinic each week day for about six weeks, posing a considerable burden to both patients and psychiatric clinics. Methods to predict who responds to the treatment could reduce this burden. Electroencephalography (EEG) has been used for decades to examine how brain activity relates to cognitive functions and psychiatric conditions. Some research has suggested that certain EEG measures provide some success at predicting response to TMS. However, many of those measures do not have predictive success when applied to new datasets, and the predictive accuracy is not high enough for clinical application. This project aims to use highly comparative machine learning methods to determine if we can extract reliable brain activity predictors of response to TMS treatment of depression, and possibly whether responses to other treatments can also be predicted. Additionally, it may be possible to apply the same analyses to reveal the mechanism of action of TMS treatment, by analysing data from before and after treatment in individuals who respond to treatment compared to individuals who do not respond.
The Project Client and Supervisors
The John Eccles institute (JEI) is a new Institute within the ANU College of Health and Medicine, created to harness transdisciplinary opportunities and engagement across the broadest conception of neuroscience—the study of the brain and mind—for research, education and public engagement across the widest possible resources within the ANU and beyond. The Monarch Research Institute is also a new institute affiliated with the ANU College of Health and Medicine, which focuses on researching novel treatments for psychiatric illnesses.
During the project, you will have a dream opportunity to be supervised by a transdisciplinary panel of excellent and committed leaders in research, education, service, and leadership. The project will be co-supervised by
- Professor Paul Fitzgerald (Director, ANU School of Medicine and Psychology, Director, Monarch Research Institute), renowned psychiatrist and a clinical and academic leader,
- Professor Hanna Suominen (Associate Director of JEI, Neuroinformatics), a computer scientist with 20 years’ of experience at forefront of bringing machine/deep learning algorithms, document analysis methods, and personalised medicine technologies to bear, and
- Dr Neil Bailey (Senior Research Fellow, ANU School of Medicine and Psychology, Head of Data Science, Monarch Research Institute), a neuroscientist with 12 years of expertise in clinical EEG research.
The First Phase—6-month Timeline
The first phase would involve identifying potential open access databases of EEG data from patients undergoing treatments for depression for analysis (we are aware of some already, but there are likely more that could be usefully analysed). It would also involve identifying machine learning software packages that will provide an optimal ability to comprehensively assess EEG data for meaningful signals that might predict treatment responses.
After the first 6 months, the project would involve using the identified datasets and software packages to determine the most meaningful EEG measures for predicting treatment responses. Following the initial analyses, attempts will be made to refine the results to obtain more accurate predictions, using for example, deep learning and reinforcement learning techniques. If the analyses are successful, the project offers the potential for objective, brain activity based treatment response predictions for psychiatric conditions. The mechanism of action analysis may also offer novel therapeutic targets for brain stimulation techniques developed by future research.
To undertake substantial work on this project, students must have a background in statistical and data analytic techniques, machine learning, and experience with processing EEG data.