At the University of Florida, researchers have cracked a diagnostic puzzle that has stumped clinicians for decades: they developed an artificial intelligence tool called AIDD—Automated Imaging Differentiation for Dementia—that can distinguish between Alzheimer's disease and dementia with Lewy bodies with near-perfect accuracy.
The distinction matters far more than medical jargon suggests. While both are forms of dementia, they present differently and require entirely different treatments. Dementia with Lewy bodies often strikes first through attention problems, drowsiness, and movement difficulties, whereas Alzheimer's typically announces itself with memory loss. Yet up to 50% of people actually living with Lewy body dementia are misdiagnosed as having Alzheimer's instead. When that happens, patients may receive treatments that actively worsen their cognitive and motor functions—a cascade of harm rooted in a simple mix-up.
Today's diagnostic approach relies on a patchwork of evaluations, tests, and brain scans rather than any single definitive measure. The confusion is understandable but costly. As dementia cases are expected to more than double by 2060, getting the diagnosis right from the start becomes urgent.
David Vaillancourt, a distinguished professor at UF's Department of Applied Physiology & Kinesiology, led the team alongside Angelos Barmpoutis and postdoctoral student Robin Chen. They analyzed 519 brain scans collected between January 2007 and March 2022 from research centers across multiple locations. The team focused on a refined dataset of 387 scans—129 from Alzheimer's patients, 129 from those with dementia with Lewy bodies, and 129 from people without disease—to train and test their AI model. Eighty percent of these scans taught the machine; the remaining 20% tested its accuracy.
The breakthrough hinges on a specialized MRI technique that detects subtle water-movement patterns in the brain, measuring extra fluid that signals cell damage and inflammation. These delicate signals, invisible to the naked eye, became readable to the AI system. The researchers published their findings in Neurology, demonstrating strong performance across multiple brain-scan comparisons.
But the real test came from the dead. To validate their tool against ground truth, researchers applied AIDD to 13 patients whose diagnoses had been confirmed through autopsy. The tool correctly identified all 13 cases—a flawless record that speaks to its potential.
"Since the therapies for Alzheimer's disease and dementia with Lewy bodies differ, developing precision biomarkers will offer better outcomes for patients," Vaillancourt said. The team performed extensive validation experiments using data from multiple scanners and imaging centers to ensure the highest standards of reliability, recognizing that this tool's value hinges entirely on its trustworthiness across different settings.
What AIDD represents is a shift from guesswork to precision—from a diagnostic process riddled with coin-flip odds to one grounded in measurable brain biology. As Alzheimer's and Brain Awareness Month shines a light on these conditions, researchers are quietly building the tools that could spare hundreds of thousands of people from misdiagnosis and the wrong treatment. The promise is simple: earlier, more accurate diagnosis. The impact could reshape how millions with dementia are cared for over the coming decades.