Machine Learning Algorithm Accurately Diagnoses Early-Stage Alzheimer’s Disease from a Single MRI Scan

Machine Learning Algorithm Accurately Diagnoses Early-Stage Alzheimer’s Disease from a Single MRI Scan

A machine learning algorithm has been developed by researchers at Imperial College London that is capable of diagnosing Alzheimer’s disease from a single magnetic resonance imaging (MRI) scan. The tool can diagnose the disease in the early stages when it would otherwise be very difficult to detect and can diagnose the disease without the need for a subject matter expert.

Alzheimer’s disease is the most common form of dementia and affects more than 500,000 people in the United Kingdom. The disease is most common in individuals over the age of 65, although Alzheimer’s disease can affect younger individuals. The disease affects thinking, can cause problems with language and problem solving, affects behavior, and causes memory loss.

There is currently no cure for Alzheimer’s disease, but getting an early diagnosis is important as treatments are available for helping patients manage the symptoms, and early diagnosis ensures Alzheimer’s disease patients can get the support and care they need and can plan for the future. A diagnosis is usually made after tests of cognitive function, memory, and brain scans, but diagnosing Alzheimer’s disease in the early stages using these methods is difficult.

The researchers developed their algorithm from one that was used for classifying cancerous tumors and trained it to look for changes in the brain. They divided the brain into 115 different regions and trained the algorithm to look at 660 different features in those brain regions and assess changes, which allowed the algorithm to accurately predict whether the patient had Alzheimer’s disease, long before there was any obvious shrinkage of the brain. According to the researchers, “For each patient, a biomarker called “Alzheimer’s Predictive Vector” (ApV) was derived using a two-stage least absolute shrinkage and selection operator (LASSO).”

The algorithm was tested using brain scans from more than 400 patients with early- or late-stage Alzheimer’s disease obtained from the Alzheimer’s Disease Neuroimaging Initiative. They also tested it on brain scans from more than 80 patients who were undergoing diagnostic tests for Alzheimer’s disease at Imperial College Healthcare NHS Trust.

The algorithm was able to correctly diagnose Alzheimer’s disease in 98% of patients and could distinguish between early- and late-stage Alzheimer’s disease in 79% of patients.  When patients visit clinics with complaints of memory problems, they often have other neurological conditions. The researchers demonstrated their tool was able to differentiate between patients with Alzheimer’s disease and other neurological conditions. The tool also identified changes in the brain in areas previously not associated with Alzheimer’s disease, such as the cerebellum and ventral diencephalon, which could open new avenues to explore in Alzheimer’s disease research.

“This method provides a biomarker able to detect an early stage of AD with a significant potential improvement of the clinical decision support system,” explained the researchers in the paper. “Our ApV is robust and repeatable across MRI scans, demonstrating its potential for applicability in clinical practice in the future.”

You can read more about the study in the paper – A predictive model using the mesoscopic architecture of the living brain to detect Alzheimer’s disease – which was recently published in Nature Communications Medicine. DOI: 10.1038/s43856-022-00133-4