Mayo Clinic researchers have developed a brand new synthetic intelligence (AI) instrument that helps clinicians determine mind exercise patterns linked to 9 kinds of dementia, together with Alzheimer’s illness, utilizing a single, broadly obtainable scan—a transformative advance in early, correct analysis.
The instrument, StateViewer, helped researchers determine the dementia kind in 88% of circumstances, in line with analysis printed on-line on June 27, 2025, in Neurology, the medical journal of the American Academy of Neurology. It additionally enabled clinicians to interpret mind scans practically twice as quick and with as much as thrice better accuracy than customary workflows. Researchers educated and examined the AI on greater than 3,600 scans, together with pictures from sufferers with dementia and other people with out cognitive impairment.
This innovation addresses a core problem in dementia care: figuring out the illness early and exactly, even when a number of circumstances are current. As new therapies emerge, well timed analysis helps match sufferers with essentially the most applicable care when it will probably have the best impression. The instrument might carry superior diagnostic assist to clinics that lack neurology experience.
The rising toll of dementia
Dementia impacts greater than 55 million folks worldwide, with practically 10 million new circumstances annually. Alzheimer’s illness, the commonest type, is now the fifth-leading explanation for loss of life globally. Diagnosing dementia sometimes requires cognitive exams, blood attracts, imaging, medical interviews and specialist referrals. Even with in depth testing, distinguishing circumstances reminiscent of Alzheimer’s, Lewy physique dementia and frontotemporal dementia stays difficult, together with for extremely skilled specialists.
StateViewer was developed underneath the path of David Jones, M.D., a Mayo Clinic neurologist and director of the Mayo Clinic Neurology Synthetic Intelligence Program.
“Each affected person who walks into my clinic carries a singular story formed by the mind’s complexity,” Dr. Jones says. “That complexity drew me to neurology and continues to drive my dedication to clearer solutions. StateViewer displays that dedication—a step towards earlier understanding, extra exact therapy and, at some point, altering the course of those ailments.”
To carry that imaginative and prescient to life, Dr. Jones labored alongside Leland Barnard, Ph.D., a knowledge scientist who leads the AI engineering behind StateViewer.
“As we had been designing StateViewer, we by no means overpassed the truth that behind each information level and mind scan was an individual dealing with a tough analysis and pressing questions,” Dr. Barnard says. “Seeing how this instrument might help physicians with real-time, exact insights and steering highlights the potential of machine studying for medical medication.”
Turning mind patterns into medical perception
The instrument analyzes a fluorodeoxyglucose positron emission tomography (FDG-PET) scan, which reveals how the mind makes use of glucose for power. It then compares the scan to a big database of scans from folks with confirmed dementia diagnoses and identifies patterns that match particular varieties, or mixtures, of dementia.
Alzheimer’s sometimes impacts reminiscence and processing areas, Lewy physique dementia includes areas tied to consideration and motion, and frontotemporal dementia alters areas liable for language and habits. StateViewer shows these patterns by means of color-coded mind maps that spotlight key areas of mind exercise, giving all clinicians, even these with out neurology coaching, a visible rationalization of what the AI sees and the way it helps the analysis.
Mayo Clinic researchers plan to broaden the instrument’s use and can proceed evaluating its efficiency in a wide range of medical settings.
Extra data: Leland Barnard et al, An FDG-PET–Primarily based Machine Studying Framework to Assist Neurologic Determination-Making in Alzheimer Illness and Associated Issues, Neurology (2025). DOI: 10.1212/WNL.0000000000213831
