Grant: Doctoral Research Award: Canada Graduate Scholarships

Epilepsy is a common and debilitating disease, characterized by recurrent seizures as the result of abnormal and excessive neuronal discharges. Its diagnosis frequently relies on the electroencephalogram (EEG). During an EEG, electrodes on the patient's scalp record the brain's electrical activity, typically for 30 minutes. Neurologists then visually examine these recordings looking for interictal epileptiform discharges (IEDs), i.e. "spiky" discharges sometimes occurring between seizures in epileptic patients. Unfortunately, a routine EEG will manage to capture these spikes in less than half of epileptic patients. In absence of IEDs, neurologists have no other means to distinguish an EEG of an epileptic patients from that of a non-epileptic patient. Recently, researchers have turned towards computers to interpret EEGs more objectively, and possibly extract subtle patterns that may be missed by the human eye. Deep learning (DL), a type of artificial intelligence, is revolutionizing our modern society. Inspired by the organization of neurons in the brain, DL allow computers to capture hidden patterns in images and other signals. Applied to EEG, DL is capable of automatically identifying seizures. Could DL detect hidden, abnormal patterns in the brain signal of epileptic patients? The objectives of the project are: 1) To extract thousands of anonymized EEG recordings from the CHUM database and review the medical files to identify which EEG are from epileptic patients and which are not; and 2) To implement a DL algorithm and train it to recognize epilepsy in seemingly normal EEGs. The impact of this project is important. Identifying new patterns of epilepsy could increase the sensitivity of the EEG in epilepsy, potentially automate its interpretation, and improve its availability to a larger percentage of the population.