Here’s how AI could aid in early Alzheimer’s detection

University of Saskatchewan researchers want to make early detection of Alzheimer’s as easy as a routine eye exam.

Neuroscientist Changiz Taghibiglou and postdoctoral fellow Sara Madranisamani are using data from retinal scans and artificial intelligence to develop a screening tool that could predict the disease decades before symptoms appear.

“This is non-invasive. There is no harm, no pain to the patient,” Taghibiglou said.

The team is working off previous research that suggests changes in the layers of the retina and optic nerve can indicate Alzheimer’s 10 to 20 years before a diagnosis, Taghibiglou said.

“Eyes are the windows of the brain because they sit directly connected to the brain (and) to the optic nerve,” he said.

The project, which includes researchers from areas in neuroscience, computer science and medicine, will use AI to train an algorithm that can use retinal images taken from a routine eye exam to detect early patterns of degeneration.

“If we train more and more and more, we will get more accurate results,” Madranisamani said.

The group wants to collect 20 years’ worth of retinal scans from at least 100 Saskatchewan patients who were recently diagnosed with Alzheimer’s. They will compare the results to scans from 100 healthy patients. That data will be used to develop the algorithm.

“AI is working in different parts of the medical (field) from the detection, prediction (and) treatment,” Madranisamani said.

“It plays a pivotal role in those areas.”

An estimated 50 million people worldwide are living with Alzheimer’s, a number that is expected to triple by 2050.

There is no cure for Alzheimer’s, but if caught early, changes to lifestyle and diet can help slow the progression of the disease.

“The majority of the patients that I see in my clinic are seniors and elderly patients who are at risk for developing this problem. So it might be a very good chance at early detection,” said ophthalmologist Dr. Ravi Nrusimhadevara, who is also a partner in the research project

The researchers are confident that an accurate algorithm can be developed within the next two years. Eventually, they hope to expand the research into a database or app that can help clinics improve patient care.

If it works, Taghibiglou said a similar approach could be applied to other diseases such as Parkinson’s.

The research recently received funding through the Saskatchewan Health Research Foundation and is in its early stages.