Wednesday, May 16, 1:15 p.m. – 2:45 p.m.
Presenter: Dr. Kiret Dhindsa, McMaster University
Topic: A brain-computer interface for communication with coma patients
Analysis of brain activity through electroencephalography (EEG) has shown that a substantial number of coma patients are conscious and aware of their surroundings. Though these patients may be able feel and hear, their condition can leave them without any means to respond to people or their environments. As such, they are unable to communicate their needs and wishes, and are even unable to express themselves to their loved ones. For these patients, even the most rudimentary mode of communication can be life-changing.
Based on our recent innovations in personalized brain-computer interfacing, we are developing a system that allows its user to answer “Yes” or “No” questions by thought alone. The system works by using deep neural networks to perform real-time analysis and interpretation of EEG, or brain waves. One model continuously scans the user’s brain activity to determine whether they intend to answer a question. This model acts as a switch, rerouting the brain signal to a second model that determines whether the answer is “Yes” or “No”. The system automatically calibrates to each user based on a set of questions for which the answers are already known, allowing the models to learn each individual’s unique patterns of brain activity.
We are currently piloting this system with healthy volunteers before deploying it with conscious coma patients at St. Joseph’s Hospital in Hamilton, Ontario. Early results show that the current system can identify the correct answer over 70% of the time, though there remain several avenues for improvement.
© 2018 Compute Ontario