ONArtificial intelligence may detect early signs of Alzheimer's disease more than six years before a patient's normal diagnosis, research suggests.
Scientists conducting a small pilot study trained a self-learning computer program to detect features in brain scans that are too subtle to humans.
The system was able to detect the onset of Alzheimer's disease in 40 patients on average more than six years before their formal diagnosis.
The American researchers trained the "deep learning algorithm" with more than 2,100 PET images (positron emission tomography) of 1,002 patients.
PET scans measure metabolic activity in the brain by monitoring the uptake of a radioactive glucose compound injected into the blood.
Research has linked the development of Alzheimer's to metabolic changes in certain parts of the brain, but these may be difficult to detect.
After examining thousands of scans, the device learned to recognize the patterns indicated for the disease.
As a test, the algorithm received a set of 40 scans from 40 patients he had never studied before. In the detection of Alzheimer's disease, on average, more than six years before the definitive diagnosis of a patient, it was 100% accurate.
Dr. Jae Ho Sohn, a member of the University of California team at San Francisco, said, "We were very pleased with the algorithm's performance, predicting every single case that progressed to Alzheimer's disease."