BU creates new AI tool to predict markers of Alzheimer’s disease

(Boston)—Researchers at Boston University Chobanian & Avedisian School of Medicine have built an artificial intelligence (AI) tool that can accurately predict key signs of Alzheimer's disease—such as the presence of sticky proteins called amyloid beta and tau—using common and less expensive tests like brain scans, memory checks and health records.

“We used data from multiple international research cohorts, allowing us to predict the presence of these sticky proteins, and even checking specific brain areas,” explains corresponding author Vijaya B. Kolachalama, PhD, FAHA, associate professor of medicine and computer science at Boston University. While popular new blood tests can somewhat detect signs of Alzheimer's, they can't reveal exactly where in the brain the issues are occurring—unlike our AI tool, which provides important location-specific detail.

Kolachalama and his team gathered information from seven different cohorts, totaling 12,185 participants, including their age, health history, memory test scores, genetic information and brain scans. They trained an AI model on this data to learn patterns that match the presence of sticky proteins seen in expensive scans and even designed the model to work if some of the information was missing. They then tested it on a separate group of people not used in training and found that the AI correctly predicted who had high amyloid or tau levels.

Kolachalama believes this tool could make checking for Alzheimer's disease easier and less costly for everyone. “The tool can help doctors quickly pick people for treatment with new drugs or to participate in research studies, thus saving time and money while reaching more patients who might not have access to costly and complicated tests. For the public, this means faster diagnoses, fewer unnecessary exams and hope for treatments that slow the disease, improving daily life for those affected and their loved ones,” he adds.

According to the researchers, this study suggests AI could also change how we stage the disease, spotting it early before symptoms get bad, which might lead to personalized plans, like custom diets or exercises to slow it down. Additionally, they feel one day this tool could impact other disorders with similar protein issues, like frontotemporal dementia, a type of brain shrinkage causing personality changes and chronic traumatic encephalopathy, brain damage from head injuries, common in athletes.

These findings appear online in the journal Nature Communications.

This project was supported by grants from the National Institute on Aging’s Artificial Intelligence and Technology Collaboratories (P30-AG073105), the American Heart Association (20SFRN35460031), Gates Ventures, and the National Institutes of Health (R01-HL159620, R01-AG062109, R01-NS142076 and R01-AG083735).

Note to editors:

 Kolachalama is a co-founder and equity holder of deepPath Inc., and CogniScreen, Inc. He also serves on the scientific advisory board of Altoida Inc. R.A. The remaining authors declare no competing interests.