Diagnosing brain diseases has long been a domain that requires a high level of expertise. Particularly for diseases like dementia or Parkinson's disease, which have similar symptoms and progress slowly, accurate differentiation has been difficult without collaboration from experienced specialists and expensive equipment.
Recently, advancements in artificial intelligence (AI) have increased the accuracy of brain imaging diagnoses while decreasing dependence on experts. The AI diagnostic tool implemented by the Mayo Clinic distinguishes between dementia and Parkinson's disease faster and more precisely than humans, fundamentally changing the landscape of medical practice.
◇Analyzing brain metabolic micro-patterns without collaboration
According to the medical journal STAT on the 5th (local time), the AI diagnostic tool 'StateViewer' developed by the largest general hospital in the United States, the Mayo Clinic, has shown results in PET diagnosis without the need for experts just four months after its introduction.
PET is a method that identifies the affected area using positrons, which have the same physical properties as electrons but opposite electric properties. After administering a radioactive pharmaceutical, it detects the points where positrons are emitted to show images of the disease area.
Degenerative brain diseases like dementia and Parkinson's disease have been diagnosed early through PET scans, but the equipment is very expensive and there is the issue of administering radioactive materials. Collaboration among various specialists such as neurology, radiology, and nuclear medicine has also been necessary, creating limitations in utilization.
StateViewer is an AI that has learned from FDG PET brain imaging data of 3,670 patients collected by the Mayo Clinic over several years. FDG PET checks how much glucose brain cells use to capture changes in brain function early. Degenerative brain diseases like Alzheimer's disease, Lewy body dementia, and Parkinson's disease show abnormalities in brain metabolic function before the disease progresses.
According to the Mayo Clinic, StateViewer automatically analyzes subtle brain metabolic patterns in FDG PET images that are difficult for humans to distinguish, providing diagnostic results within minutes. Lewy body dementia, which accounts for 20% of dementia patients, is often missed even by specialists with just the images, but the AI accurately detects these subtle patterns.
Dr. David Jones, who oversees the AI neurology program at the Mayo Clinic, noted, "AI helps physicians make independent judgments without collaboration, enhancing individual judgment capabilities and drastically reducing diagnosis times."
◇AI lowers dependence on specialists and expenses
The AI diagnostic technology has been credited with lowering the financial and professional barriers to PET exams. It contributes not only to early diagnosis but also to more accurate treatment direction setting for individual patients. The recent Alzheimer's treatment is effective for patients in early stages. As the importance of early diagnosis increases with the release of new dementia drugs, the likelihood of StateViewer being utilized in other hospitals is rising.
Recently, AI has evolved from simply analyzing medical images to a method that comprehensively analyzes patient information for diagnoses. Microsoft announced an AI doctor at the end of June that can converse with patients like a physician and perform necessary tests for diagnosis. In a comparison of actual diagnostic cases, the AI demonstrated a diagnostic accuracy of 85.5%, surpassing human physicians by 20%.
Experts believe that such comprehensive judgment AI is necessary for diagnosing brain diseases. Zoe Kourtzi, a professor of cognitive neuroscience at the University of Cambridge in the UK, stated, "To accurately assess brain diseases, various data such as brain images, blood tests, genetic information, and medical history need to be synthesized. It's practically impossible to make accurate diagnoses based solely on images, and we need an AI model that can interpret these comprehensively."
AI diagnosing brain diseases is emerging as a new growth axis in the global healthcare market. In Korea, efforts to develop brain diagnostic AI are accelerating. Notably, Neurophet has developed 'Neurophet Aqua', which analyzes brain nerve images, and 'Neurophet Scale PET', which analyzes PET images, as well as an AI called 'Neurophet Aqua AD' that manages prescriptions, effects, and side effects of dementia medications.
Additionally, VUNO has developed an AI called 'DeepBrain' that analyzes brain MRI images to diagnose Alzheimer's disease, while DEEPNOID has developed an AI called 'DeepNeuro' that analyzes brain magnetic resonance angiography (MRA) images. This AI read MRA images 66 minutes faster than radiology residents and 60 minutes faster than radiology specialists in clinical trials involving 330 patients.
A representative from a domestic AI diagnostic company stated, "The clinical results of StateViewer show that AI has the potential to significantly reduce dependence on specialists, going beyond being a simple assistive tool. As AI changes the paradigm of early diagnosis in brain diseases, attention is focused on how the combination of human expertise and machine learning will alter the landscape of healthcare in the future."