Researchers at Worcester Polytechnic Institute have developed an artificial intelligence tool that analyzes brain MRI scans to predict Alzheimer’s disease with 92.87% accuracy. The study highlights subtle anatomical changes, such as brain volume loss, that vary by age and sex, offering potential for earlier diagnosis.
Challenges in Early Alzheimer’s Detection
Alzheimer’s disease, a progressive neurodegenerative disorder, affects an estimated 6.9 million Americans aged 65 and older. It damages neurons, causing cell death, brain tissue loss, and decline in cognitive functions like memory and learning. Early symptoms often mimic normal aging, complicating diagnosis.
Machine learning overcomes these hurdles by processing vast MRI data to detect minute changes invisible to the human eye. “Machine-learning technologies can analyze large amounts of data from scans to identify subtle changes and accurately predict Alzheimer’s disease and related cognitive states,” states Benjamin Nephew, assistant research professor in the Department of Biology and Biotechnology. “This advance informs Alzheimer’s research and may enable doctors to diagnose and treat the disease earlier and more effectively.”
Study Methodology and Results
The research team, including PhD student Senbao Lu and recent MS graduate Bhaavin Jogeshwar, examined 815 MRI scans from the Alzheimer’s Disease Neuroimaging Initiative database. These scans represented individuals aged 69 to 84 with normal cognition, mild cognitive impairment (MCI), or Alzheimer’s disease.
Using machine learning, they measured volumes in 95 brain regions and applied an algorithm to differentiate healthy brains from those with MCI or Alzheimer’s. The model achieved 92.87% accuracy across groups.
Key predictors included volume loss in the hippocampus (crucial for memory and learning), amygdala (regulates emotions), and entorhinal cortex (early Alzheimer’s target involved in memory, navigation, and perception).
Age- and Sex-Specific Brain Changes
Analysis revealed patterns differing by demographics. In the youngest group (ages 69-76), both males and females showed right hippocampus volume loss, underscoring its role in early detection.
Sex differences stood out: females experienced loss in the left middle temporal cortex, linked to language, memory, and visual perception, while males showed notable reduction in the right entorhinal cortex. “The degree of these differences was surprising and may relate to interactions between Alzheimer’s progression and sex hormone changes,” Nephew notes. Declining estrogen in women and testosterone in men could elevate risk.
“The critical challenge is building a generalizable machine-learning model that captures differences between healthy brains and those with MCI or Alzheimer’s,” Nephew explains. “A generalizable model means the biomarkers are universal across patients.”
Future Directions and Interdisciplinary Impact
The team now explores deep learning models and factors like diabetes that influence brain health and Alzheimer’s. The work draws students from biology, neuroscience, psychology, computer science, and bioinformatics.
“This research exemplifies the strength of interdisciplinary, computational neuroscience,” Nephew says. “The brain is complex, demanding broad approaches to understand, predict, and treat its diseases.”
The findings appear in Neuroscience (DOI: 10.1016/j.neuroscience.2025.12.030).
