Computers can 'learn' to identify patterns in Alzheimer's patients' brains

Computers might be able to detect early signs of Alzheimer’s that their human operators aren’t even aware exist, according to a new study in Radiology.

One type of artificial intelligence, called machine learning, is programmed to recognize emerging patterns as it’s fed new data, without having to be “taught” exactly what it’s looking for. Researchers examined what that could mean for detecting Alzheimer’s disease in its early stages, when human doctors aren’t yet totally sure what they’re looking for.

Between 2010 and 2012, 260 health subjects and patients with dementia underwent arterial spin labeling MRIs to measure blood flow to different areas of the brain. The machine learning program was able to identify those patients with the most severe dementia diagnoses.

Eventually, machine learning could mean that computers pick up on patterns in the brain scans of patients who go on to develop Alzheimer’s disease, and the computers will learn what to look for in the MRIs of future patients’ brains, even when doctors don’t see the patterns themselves. Already, doctors (with help from the computer program) were able to predict progression of Alzheimer’s or dementia in a given patient during the study time with between 82 and 90 percent accuracy.

With Alzheimer’s the fifth-leading cause of death for adults older than 65, the ultimate goal is for both the machines and the physicians to learn how to identify the disease early enough to maximize treatment options.  

"With standard diagnostic MRI, we can see advanced Alzheimer's disease, such as atrophy of the hippocampus," lead researcher Alle Meije Wink, PhD, said in a statement. "But at that point, the brain tissue is gone and there's no way to restore it. It would be helpful to detect and diagnose the disease before it's too late."