# WashU AI Classifier Differentiates Major Dementia Types > Source: > Published: 2026-05-30 17:23:16.614081+00:00 # WashU AI Classifier Differentiates Major Dementia Types Researchers at Washington University School of Medicine developed an AI classifier that distinguishes between four common causes of dementia-Alzheimer's disease, Parkinson's disease, frontotemporal dementia, and dementia with Lewy bodies-as well as healthy brain aging, reporting over **90% accuracy** (WashU Medicine; Futurity). The team selected a panel of **15 proteins** measured in blood and trained and tested the model on blood-protein data from **more than 3,200 individuals**, per Futurity. The classifier can also detect coexisting disease processes, a frequent clinical challenge. The findings appear in the journal **Alzheimer's & Dementia**, and senior author Carlos Cruchaga is quoted on the limits of single-diagnosis approaches (WashU Medicine). ### What happened Researchers at **Washington University School of Medicine** developed an AI-based classifier that the team reports can distinguish among **Alzheimer's disease**, **Parkinson's disease**, **frontotemporal dementia**, **dementia with Lewy bodies**, and **healthy brain aging**, with reported accuracy above **90%** (WashU Medicine; Futurity). The work appears in **Alzheimer's & Dementia** (WashU Medicine). The researchers say the model can identify when more than one neurodegenerative process is present in the same patient, a common and clinically challenging situation (WashU Medicine; Futurity). "Right now, many patients get labeled with a single diagnosis of, say, Alzheimer's or Parkinson's, but in reality their brains often show a mixture of disease injuries," said Carlos Cruchaga, per WashU Medicine. ### Technical details The team selected a set of **15 proteins** measured in blood that the authors describe as reflecting Alzheimer's pathology, synapse and nerve damage, and inflammation (Futurity; WashU Medicine). Per Futurity, the classifier was trained and tested on blood-protein data from **more than 3,200 individuals** collected through a large research cohort. The paper and institutional release do not publish the classifier architecture details in the cited summaries; the sources emphasize the protein panel and cohort scale. ### Editorial analysis - technical context Blood-based biomarker panels combined with machine learning are an accelerating pattern in neurodegenerative research. Industry-pattern observations: such approaches can compress complex neuropathology into low-dimensional signals useful for screening and triage, but they commonly face reproducibility, preanalytical variability, and cohort-bias challenges that require multi-site validation and assay standardization before clinical deployment. ### Context and significance Industry context: A scalable, minimally invasive test that separates multiple dementia etiologies would affect clinical trial enrollment, longitudinal monitoring, and differential-treatment decisions in research settings. Observers will note that demonstration of diagnostic accuracy in retrospective or convenience cohorts is an early step; regulators and clinicians typically require prospective validation and performance replication across diverse populations. ### What to watch Watch for independent replication cohorts, prospective clinical trials, methods and code disclosure in the journal paper, analytical validation of the 15-protein assay, and any reported steps toward clinical-grade assay standardization or commercial partnerships. Washington University has not been quoted in the sources as announcing regulatory filing timelines (WashU Medicine; Futurity). ## Scoring Rationale Notable applied-AI result with direct relevance to clinical diagnostics and biomarker research. The work demonstrates promising accuracy on a large cohort, but broader validation, regulatory review, and standardization are necessary before clinical adoption. Practice interview problems based on real data 1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with. [Try 250 free problems](/problems)