
A team of researchers has recently developed an artificial intelligence (AI) tool capable of predicting Alzheimer’s disease years before actual diagnosis. The scientists, from the University of California in San Francisco, used taught an AI system to analyze brain scans to detect these early indicative signs of the disease. They are confident that after further validation, their AI system could have great impact in early detection of Alzheimer’s, allotting patients more time to let treatments effectively slow the disease’s progression.
Using over 2,109 positron-emission tomography (PET) images of 1,002 patients’ brains, the researchers trained the AI system in a deep learning algorithm-driven process. This algorithm was then tested on PET images of 40 other patients’ brains to test its efficacy. They found that the AI system was able to accurately predict which patients would eventually receive Alzheimer’s diagnosis more than 6 years in advance.
Published in Radiology, the researchers describe that their system yielded “82 percent specificity at 100 percent sensitivity, an average of 75.8 months prior to the final diagnosis.”
READ: Bill Gates’ New Investment Seeks to Revolutionize Alzheimer’s Testing
During Alzheimer’s progression, the metabolism of glucose in the patient’s brain is altered. This abnormality appears on PET imaging that tracks the brains metabolism of 18F-fluorodeoxyglucose (FDG), a radioactive form of glucose. Programming the algorithm to look for specific aspects of these FDG PET images, researchers were able to “train” it to detect early metabolic changes associated with Alzheimer’s. This deep learning process involves complex AI training through examples, similar to how humans learn. Data from the Alzheimer’s Disease Neuroimaging initiative was used to teach this AI system, in addition to the 2,000+ images mentioned prior.
“We were very pleased with the algorithm’s performance,” said co-author Dr. Jae Ho Sohn, member of the University of California in San Francisco’s radiology and biomedical imaging departments. “It was able to predict every single case that advanced to Alzheimer’s disease.”
READ: The WAYBACK Project: VR Helping People with Alzheimer’s
Alzheimer’s disease is estimated by the Alzheimer’s Association to affect roughly 5.7 million US citizens, and they predict this number will increase to almost 14 million by 2050. The neurodegenerative disease takes a toll on essential skills such as cognition and vision and can severely impede on daily activities. The atrophy of tissue in the brain only worsens with time when untreated, therefore this early detection AI system could bolster Alzheimer’s care immensely. Not only would this help sufferers from the disease and their families, but it would greatly reduce costs associated with long term Alzheimer’s care.
Though these results are profound, Dr. Sohn cautions readers that the sample size was relatively small and that further validation of the system is needed.
Dr. Sohn cautions that the study was small and that the findings now need to undergo validation. Expanding clinical trials of the AI system will entail using larger sets of data, and more images taken from patients of various locations. The research team also plans to implement other patterns of recognition into their algorithm. Study co-author and professor in Radiology and Biomedical Imaging, Youngho Seo, notes that patterns such as protein buildup in the brain also indicates Alzheimer’s.
He claims: “If FDG PET with [artificial intelligence] can predict Alzheimer’s disease this early, beta-amyloid plaque and tau protein PET imaging can possibly add another dimension of important predictive power.”
A team of researchers has recently developed an artificial intelligence (AI) tool capable of predicting Alzheimer’s disease years before actual diagnosis. The scientists, from the University of California in San Francisco, used taught an AI system to analyze brain scans to detect these early indicative signs of the disease. They are confident that after further validation, their AI system could have great impact in early detection of Alzheimer’s, allotting patients more time to let treatments effectively slow the disease’s progression.
Using over 2,109 positron-emission tomography (PET) images of 1,002 patients’ brains, the researchers trained the AI system in a deep learning algorithm-driven process. This algorithm was then tested on PET images of 40 other patients’ brains to test its efficacy. They found that the AI system was able to accurately predict which patients would eventually receive Alzheimer’s diagnosis more than 6 years in advance.
READ: VR Simulating Dementia Makes Youth Sympathize with Elders
Published in Radiology, the researchers describe that their system yielded “82 percent specificity at 100 percent sensitivity, an average of 75.8 months prior to the final diagnosis.”
During Alzheimer’s progression, the metabolism of glucose in the patient’s brain is altered. This abnormality appears on PET imaging that tracks the brains metabolism of 18F-fluorodeoxyglucose (FDG), a radioactive form of glucose. Programming the algorithm to look for specific aspects of these FDG PET images, researchers were able to “train” it to detect early metabolic changes associated with Alzheimer’s. This deep learning process involves complex AI training through examples, similar to how humans learn. Data from the Alzheimer’s Disease Neuroimaging initiative was used to teach this AI system, in addition to the 2,000+ images mentioned prior.
“We were very pleased with the algorithm’s performance,” said co-author Dr. Jae Ho Sohn, member of the University of California in San Francisco’s radiology and biomedical imaging departments. “It was able to predict every single case that advanced to Alzheimer’s disease.”
Alzheimer’s disease is estimated by the Alzheimer’s Association to affect roughly 5.7 million US citizens, and they predict this number will increase to almost 14 million by 2050. The neurodegenerative disease takes a toll on essential skills such as cognition and vision and can severely impede on daily activities. The atrophy of tissue in the brain only worsens with time when untreated, therefore this early detection AI system could bolster Alzheimer’s care immensely. Not only would this help sufferers from the disease and their families, but it would greatly reduce costs associated with long term Alzheimer’s care.
Though these results are profound, Dr. Sohn cautions readers that the sample size was relatively small and that further validation of the system is needed.
Dr. Sohn cautions that the study was small and that the findings now need to undergo validation. Expanding clinical trials of the AI system will entail using larger sets of data, and more images taken from patients of various locations. The research team also plans to implement other patterns of recognition into their algorithm. Study co-author and professor in Radiology and Biomedical Imaging, Youngho Seo, notes that patterns such as protein buildup in the brain also indicates Alzheimer’s.
He claims: “If FDG PET with [artificial intelligence] can predict Alzheimer’s disease this early, beta-amyloid plaque and tau protein PET imaging can possibly add another dimension of important predictive power.”
Researchers train #AI to spot Alzheimers disease ahead of diagnosis https://t.co/lx8LeAs66k#ArtificialIntelligence #Healthcare #Alzheimers #Diagnosis pic.twitter.com/NYN3JAyVVg
— Steve Kilpatrick (@SteveAtLogikk) November 7, 2018
Source: Medical News Today