Developing machine learning tools for routine clinical EEG

The proposed project is part of a large-scale collaboration between SFU’s Behavioral and Cognitive Neuroscience Institute (BCNI) and Fraser Health Authority (FHA) in the domain of AI applied to clinical electroencephalographic (EEG) scans recorded and evaluated in the process of diagnostic workup in FHA’s public hospitals (n = 40’000). The key goal of the SFU/FHA collaboration is to automate the process of EEG reporting by building a decision support system. Conventional review of EEG relies on neurologists to visually inspect complex, noisy, high-dimensional digital data. Such an approach is slow, not fully reliable, and suboptimal. There is considerable variability in EEG interpretation, and this variability is affected by specific reader characteristics. Given these challenges, clinical reporting of EEG is highly suitable for machine-driven automation. The focus of the proposed project at a smaller scale is on predicting diagnoses (codes for most responsible diagnosis, secondary diagnosis, etc, according to International Statistical Classification of Diseases ICD-10) from in-patient EEG scans. We aim to address the following question: what current deep learning approaches can be used to reach state-of-the-art performance for EEG classification based on diagnostic codes.

Faculty Supervisor:

Vasily Vakorin

Student:

Partner:

Taras Shevchenko National University of Kyiv;Bogomolets National Medical University

Discipline:

Computer science

Sector:

Education

University:

Simon Fraser University

Program: