BOSTON — One of the many obstacles scientists who study Alzheimer’s disease have been struggling to overcome is finding a standard method of accurately predicting the likelihood a person will be diagnosed with the condition. There is great deal of variance between the methods used to diagnose Alzheimer’s and one’s risk at different healthcare facilities. Now, an international research team led by Boston University researchers created an artificial intelligence computer algorithm that may solve this problem.
The algorithm can accurately predict the risk of Alzheimer’s disease or diagnose the disease using a magnetic resonance imaging (MRI) scan, results of tests that measure cognitive impairments, and basic information about age and gender.
“If computers can accurately detect debilitating conditions such as Alzheimer’s disease using readily available data such as a brain MRI scan, then such technologies have a wide-reaching potential, especially in resource-limited settings,” explains corresponding author Vijaya B. Kolachalama, PhD, assistant professor of medicine at Boston University School of Medicine, in a press release.
Researchers created their algorithm using a machine-learning method called deep learning. Using this method, the algorithm gradually learns “predictors” of Alzheimer’s disease and disease risk and can improve its predictors as it learns more.
The research team trained their algorithm on a subset of their data, and then validated its performance on the rest of their data. They had access to MRI scans, clinical information and demographics of people with Alzheimer’s disease and people with normal cognition gathered by four separate Alzheimer’s research centers. The algorithm was trained on one group of data that contained the information of 417 patients and tested on the other three groups which have just over 1,000 combined patients.
A group of 11 expert neurologists were also tested on a subset of the cases the algorithm was tested on. It turns out the computer algorithm performed slightly better than the neurologists in predicting the condition of the disease in the cases they were presented.
The researchers also demonstrated that the brain regions the computer algorithm identified as Alzheimer’s “predictors” had a diseased neuropathology in several of the autopsy reports of patients who had passed away.
“Not only can we accurately predict the risk of Alzheimer’s disease, but this algorithm can generate interpretable and intuitive visualizations of individual Alzheimer’s disease risk en route to accurate diagnosis,” says Kolachalama.
Researchers hope that their success with a deep learning algorithm will inspire scientists to try using this machine learning method to produce predictive models of other degenerative diseases.
Dr. Kolachalama concludes with some words of hope for the new algorithm. “If we have accurate tools to predict the risk of Alzheimer’s disease (such as the one we developed), that are readily available and which can use routinely available data such as a brain MRI scan, then they have the potential to assist clinical practice, especially in memory clinics.”
The study is published in Brain.