BioE PhD Proposal

Samuel H. Waters 

February 10th, 2021, 1:00 PM 

Link: https://bluejeans.com/925016567

Advisor

Gari Clifford, DPhil BME, Georgia Institute of Technology and Emory University 

Committee Members

Eva Dyer, Ph.D.

BME, Georgia Institute of Technology

Hua Wang, Ph.D.

ECE, Georgia Institute of Technology

Reza Sameni, Ph.D.

Emory University

Thad Starner, Ph.D.

ECE, Georgia Institute of Technology

Automated Diagnostic Tools Using Reduced Channel and Wearable EEG

EEG is a widely used tool for diagnosing a number of disorders including narcolepsy, epilepsy and sleep apnea, and can potentially aid mild cognitive impairment (MCI) diagnosis. However, it is sometimes necessary for patients to undergo lengthy in-hospital recordings which must be manually examined in full by human clinicians, such as when undergoing seizure monitoring or polysomnography, which is extremely time consuming for clinicians and requires the usage of limited resources such as hospital beds. The overhead time, resource requirements, and inconvenience to the patient of manually examined in-hospital EEG is particularly challenging for any form of long-term monitoring. Wearable at-home EEG devices which perform tasks such as sleep staging and seizure detection automatically are a possible solution, however there has been little research on automation using wearable sensors. The objective of this research is twofold: first, a combination of deep and statistical learning models will be trained to perform sleep staging, seizure detection and MCI diagnosis using large (over 100 patients), publicly available datasets using only channels or data which is also available on wearable devices. Secondly, a dataset of EEG recordings using wearable devices will be collected and used for fine-tuning models which were pre-trained on larger datasets via transfer learning. Such tools can be used to greatly simplify the diagnosis and monitoring of a number of neurological disorders.