Artificial intelligence might be able to treat various brain disorders

A work in progress

Liu mentioned that the neural implant would turn the neural networks on and off like a switch, or like restarting a computer. He also stated the complexity of the research project, noting that it won’t be as simple as it sounds, and that researchers are still trying to comprehend the complexity of the project. “Scientists don’t fully understand how it works yet,” said Liu, who is also part of the neurotechnology center CRANIA, a collaboration between the University of Toronto and the University Health Network.

The collective network and university joins neuroscientists, data and material scientists and clinicians, with the objective of improving brain health and creating alternative treatments. The neural implant project was created by Liu and his team as a prospective alternative treatment option for clients who might not respond well to current medications. They see the potential in using AI as a future effective treatment option while minimizing adverse reactions to excessive stimulation in the brain.

The research team calls the technology CMOS, which stands for complementary metal-oxide semiconductor. It allows them to reduce the device’s size and power utilization. This would, in turn, reduce any risks connected with the neural implant’s surgical procedure and long-standing usage.

In order to create the best prototype for the neural implant, they’ve tried various strategies and techniques as well. “We’ve developed many new microelectronic design techniques, such as high precision electrical stimulation with charge balancing,” Liu said.

Using deep learning

Researchers used a type of AI called deep learning (DL), a type of machine learning that utilizes artificial neural networks. DL uses a compilation of algorithms that “learn” and extract deep-level information when given new data. DL can also identify hidden biomarkers — the measurement or indication of disease — often neglected in traditional methods.

This is helpful for researchers because they can choose when to activate the neural implants based on biomarkers and not have to guess or use the stimulation continuously. “Most existing implants produce electrical stimulation at a constant rate, regardless of the patient’s condition,” Liu stated. “With DL, we can activate the neural implants at the optimal time and only when necessary.”

However, one caveat mentioned is the computational cost. Liu stated, for example, a neural implant can’t fail if it loses telecommunication service, such as when a patient goes in an elevator or an airplane. The computational cost of deep learning models would make it a challenge to incorporate such technology. In order to reduce computational costs, the research team created methods for training the models based solely on each patient’s condition.

The future of neural implants

The study on deep learning for neural implants being used to detect seizures was published in the Journal of Neural Engineering, and Liu wants to expand on the research. He said that his team’s work can be used in a broad range of clinical applications and medical procedures beyond epileptic seizures.

Liu wants to use the technology for a variety of brain disorders, which affect nearly one billion people across the globe. Along with studying the effects on epilepsy and Parkinson’s disease, he hopes new therapy treatments can be created for patients with dementia, chronic pain, Alzheimer’s disease and depression.