AI Tool Identifies Dementia Types With 88% Accuracy

Researchers at Mayo Clinic have developed a new artificial intelligence tool that may streamline dementia diagnosis by identifying characteristic brain activity patterns in a single fluorodeoxyglucose positron emission tomography (FDG-PET) scan. The tool, called StateViewer, was able to classify nine types of dementia with 88% accuracy, including Alzheimer’s disease, according to findings published on June 27, 2025, in Neurology.

The team tested the model on more than 3,600 brain scans. These included both individuals with confirmed dementia and people without cognitive impairments. In comparative evaluations, StateViewer helped clinicians reach diagnoses nearly twice as quickly and with up to three times greater accuracy than traditional workflows.

Supporting earlier, more precise care

StateViewer was designed to address one of the key challenges in dementia care: distinguishing between multiple overlapping or coexisting conditions early in the disease course. Diagnostic delays remain common, in part due to the complexity of neurodegenerative symptoms and the need for specialist input.

Current diagnostic processes often require a combination of neuropsychological testing, bloodwork, interviews and imaging, followed by evaluation from neurology experts. StateViewer aims to simplify this process, providing support even in clinics where neurology expertise may not be available.

Mayo Clinic researchers suggest that by enabling more timely diagnoses, the tool could help guide treatment strategies earlier, when interventions may have a greater effect.

How the system works

StateViewer analyzes FDG-PET scans, a standard imaging technique that reveals how the brain consumes glucose, an essential energy source. The system compares a patient’s scan against a reference database of images from individuals with confirmed diagnoses of Alzheimer’s disease, Lewy body dementia, frontotemporal dementia and other conditions.

Different forms of dementia are associated with distinct regional patterns of brain activity. Alzheimer’s disease is typically linked to changes in memory and processing centers, while Lewy body dementia affects attention and motor areas. Frontotemporal dementia is associated with activity changes in regions that regulate language and behavior.

StateViewer outputs a color-coded brain map that highlights regions showing atypical glucose metabolism. This visual format can assist clinicians, including those without specialized training in neurology, in interpreting the scan and understanding how the AI reached its classification.

"Every patient who walks into my clinic carries a unique story shaped by the brain's complexity. That complexity drew me to neurology and continues to drive my commitment to clearer answers. StateViewer reflects that commitment — a step toward earlier understanding, more precise treatment and, one day, changing the course of these diseases." 

Dr. David Jones.

Continued development and implementation

The project was led by David Jones, MD, director of the Mayo Clinic Neurology Artificial Intelligence Program, with AI engineering support from data scientist Leland Barnard, PhD. The team emphasized the importance of patient-centered design, ensuring that the system’s outputs are transparent and clinically interpretable.

Researchers plan to expand StateViewer’s use in broader clinical settings and to continue evaluating its effectiveness across diverse populations. This may include further refinement of its ability to distinguish mixed pathologies and its integration into routine diagnostic workflows.

Reference: Barnard L, Botha H, Corriveau-Lecavalier N, et al. An FDG-PET–based machine learning framework to support neurologic decision-making in Alzheimer disease and related disorders. Neurology. 2025;105(2):e213831. doi: 10.1212/WNL.0000000000213831

This article has been republished from the following materials. Note: material may have been edited for length and content. For further information, please contact the cited source. Our press release publishing policy can be accessed here.

This content includes text that has been generated with the assistance of AI. Technology Networks' AI policy can be found here.