by Deirdre Kelly
Biophysicist Joel Zylberberg runs a research lab at York University that is applying machine learning to the study of Parkinson’s disease, a progressive neurodegenerative disorder of the central nervous system affecting one in every 500 Canadians. Symptoms include tremors, impaired speech, dementia, and sleep and respiratory irregularities.
To date, there is no cure for Parkinson’s, though Zylberberg and his team at the University’s Centre for Vision Research are developing treatments to better manage the symptoms of this disease.
Research into vision systems in animals and humans, primarily the stimuli that create visual representations in the brain, has led to the creation of artificial neural networks that can detect behavioural states in patients, in particular sleep patterns affected by the disease.
A.I. algorithms have previously been used for diagnostic purposes. But with this development, Zylberberg – a Tier 2 Canada Research Chair in Computational Neuroscience and assistant professor in the Department of Physics and Astronomy at York – goes further by developing an implant that uses deep brain stimulation to treat motor fluctuations in people with the disease.
“Our advance here is to change the stimulation setting based on the patient’s behavioural state,” he says, “and we do this by installing our sleep stage detection algorithm into the controller for the brain implant.”
It works like this: the artificial neural networks developed in Zylberberg’s lab take in brain patterns recorded by electrodes inserted into the subthalamic nucleus in the basal ganglia, the part of our anatomy that regulates the body’s motor system. These electrodes not only record brain activity, they stimulate the brain with a pulsed current delivered to the nerve cells responsible for relaying the messages that plan and control movement.
The patent on Zylberberg’s invention is now pending.
“We found that our algorithms do a good job of predicting sleep state, which allows for a targeted stimulation of the brain beneficial to the patient. The next stage is to revise and improve on these patient algorithms using the new patient data being collected in clinical trials.”
Those trials, which include patient surgeries and data collection, are being managed through Zylberberg’s academic partners at Stanford, the University of Nebraska and the University of Colorado. Nebraska is the lead institution. But York (through Zylberberg) is spearheading the software side of the project promising benefits to patients.
Says Zylberberg, “We anticipate that the new implants we are developing, which adapt in real-time to the patient’s state, will reduce the disease burden and improve the quality of life for those suffering from Parkinson’s disease.” ■