Researchers at Washington University School of Medicine in St. Louis have developed an artificial intelligence-powered blood test that can distinguish between several major neurodegenerative diseases with an overall accuracy of 92.3 percent. The technology could help address one of the most persistent challenges in dementia care: accurately identifying the underlying cause of cognitive decline.
The study, published in Alzheimer’s & Dementia, demonstrates how a simple blood sample combined with AI analysis may support earlier and more precise diagnosis of Alzheimer’s disease, Parkinson’s disease, frontotemporal dementia and dementia with Lewy bodies. Importantly, the system can also detect when multiple disease processes are present in the same patient.
Complexity of dementia
Diagnosing dementia-related disorders remains difficult because symptoms often overlap and different neurodegenerative diseases frequently occur simultaneously. As a result, patients are commonly assigned a single diagnosis, even when multiple pathologies are contributing to their condition.
According to senior author Carlos Cruchaga, Professor of Psychiatry at Washington University School of Medicine, current diagnostic tools are not designed to capture this biological complexity. “Many patients receive a diagnosis of Alzheimer’s or Parkinson’s disease, while their brains may actually show evidence of several neurodegenerative processes,” he explained. “For precision medicine, we need tools that can identify the full spectrum of disease occurring in an individual.”
The newly developed classifier was designed with this challenge in mind, aiming to provide a broader picture of brain health rather than a simple positive-or-negative result for a single disorder.
Blood biomarkers
To develop the test, researchers identified 15 blood-based proteins associated with neurodegenerative disease. These biomarkers included established indicators of Alzheimer’s pathology as well as proteins linked to inflammation, neuronal injury and synaptic dysfunction.
The AI model was trained and evaluated using blood samples from more than 3,200 individuals enrolled through Washington University’s Alzheimer Disease Research Center and Movement Disorders program. The dataset included patients diagnosed with Alzheimer’s disease, Parkinson’s disease, frontotemporal dementia, dementia with Lewy bodies and cognitively healthy controls.
Researchers then validated the model using a separate cohort of 225 individuals who had undergone detailed cognitive assessments during life and neuropathological examination after death. This allowed the team to compare AI predictions directly with the actual disease burden found in brain tissue. The classifier achieved an overall diagnostic accuracy of 92.3 percent and showed strong agreement with both clinical diagnoses and post-mortem findings.
Detecting mixed pathology
One of the most notable findings was the model’s ability to identify mixed neurodegenerative pathology, a common but clinically challenging phenomenon. For example, the system detected Alzheimer’s-related biological changes in some patients who had been diagnosed primarily with Parkinson’s disease during life but later developed dementia. It also performed well in cases where diagnoses were uncertain, including individuals with mild cognitive impairment and ambiguous neurological symptoms.
Researchers found that predictions of Alzheimer’s disease closely reflected the actual burden of amyloid plaques observed during autopsy, suggesting that the test captures underlying disease processes even before a definitive clinical diagnosis is established.
Routine clinical use
Although the blood test is not yet ready for routine clinical use, researchers believe it could eventually play an important role in both research and patient care. Additional studies involving larger and more diverse populations will be needed to confirm the findings and evaluate the model’s ability to predict disease progression over time.
If validated, the technology could help physicians determine which patients require specialist referral, additional diagnostic testing or targeted treatment. It could also support the recruitment of appropriate participants for clinical trials aimed at specific neurodegenerative pathways.
As blood-based diagnostics continue to advance, tools such as this AI-driven classifier may help shift dementia care toward earlier, more accurate and more personalized diagnosis, an important step as healthcare systems prepare for a growing global burden of neurodegenerative disease.
Mayo Clinic solution
Last year, research by the Mayo Clinic showed that another new blood test can detect the presence of Alzheimer's in people with cognitive impairments with 95% accuracy in an outpatient setting. The findings have been published in Alzheimer's & Dementia and highlight the rise of innovative diagnostics in elderly care.
That particular blood test measures the presence of two biomarkers in blood plasma: Aβ42/40 and p-tau217. These are proteins associated with the characteristic amyloid plaques in Alzheimer's disease. The p-tau217 biomarker proved to be particularly reliable in this study: elevated concentrations of this biomarker were found in 95% of patients with Alzheimer's-related memory problems.
The test was administered to more than 500 patients at the Mayo Clinic Florida Memory Clinic, ranging in age from 32 to 89. The participants had various forms of memory impairment, including typical and atypical Alzheimer's, Lewy body dementia, and vascular dementia. In 56% of cases, Alzheimer's was identified as the main cause of the symptoms.
