Researchers at Washington University School of Medicine in St. Louis have cracked open one of medicine's most frustrating puzzles: telling apart the tangle of brain diseases that cause dementia. A new AI tool trained on blood proteins can now distinguish among four major neurodegenerative conditions—Alzheimer's disease, Parkinson's disease, frontotemporal dementia, and dementia with Lewy bodies—while also detecting when a patient's brain is ravaged by multiple diseases at once, a reality that affects many people but has been nearly impossible to pinpoint with existing tools.
The challenge is deeply human. Millions of people living with dementia never get a clear diagnosis because these diseases mimic one another, overlap, and resist the blunt instruments medicine has traditionally relied on. A patient labeled with "Parkinson's" might actually be experiencing Parkinson's and Alzheimer's pathology simultaneously—but they'd never know it, and their treatment would be incomplete. That's where Carlos Cruchaga and his team at WashU Medicine's NeuroGenomics and Informatics Center intervened. Their goal was elegantly simple: build a test that didn't just say yes or no to a single disease, but instead revealed the full spectrum of neurodegeneration happening in a person's brain.
The team selected 15 blood proteins that reflect neurodegenerative damage in the brain—including well-validated markers of Alzheimer's pathology alongside proteins involved in nerve damage and inflammation. They trained their AI classifier on blood protein data from more than 3,200 individuals, including those with confirmed diagnoses of Alzheimer's, Parkinson's, frontotemporal dementia, dementia with Lewy bodies, and cognitively normal controls. The data came from the Charles F. and Joanne Knight Alzheimer Disease Research Center and WashU Medicine's Section of Movement Disorders, giving the model a rich foundation of clinical and biological information.
The real test came when researchers verified their classifier on a separate group of 225 individuals whose brains had been examined at autopsy after cognitive evaluation during life. The results were striking: the tool achieved 92.3% overall diagnostic accuracy, correctly identifying cases of single neurodegenerative diseases while also demonstrating remarkable ability to detect mixed pathology. In people who'd had uncertain or evolving diagnoses—including those with mild cognitive impairment—the model's predictions for Alzheimer's pathology closely matched the actual amyloid plaques found in brain tissue after death. Perhaps most tellingly, the tool identified Alzheimer's-like biological changes in patients who carried a Parkinson's diagnosis during life but later developed dementia, proving its capacity to see past clinical labels to the true underlying biology.
What makes this breakthrough especially meaningful is its simplicity. A blood draw is inexpensive and noninvasive, putting genuine diagnostic clarity within reach for millions of people who might never undergo a brain scan or autopsy. For patients and their families navigating the fog of dementia, understanding whether one or multiple diseases are at work fundamentally changes the conversation around treatment. "Our goal was to build a test that doesn't just say 'yes' or 'no' to one disease but instead gives an indication of all the major neurodegenerative diseases happening in that person," Cruchaga explained. "That's what you really need for precision diagnosis and, ultimately, precision treatment." The test is expected to support not just early diagnosis, but ongoing monitoring and personalized care—transforming how doctors think about dementia in the years ahead.