Six years before diagnosed by doctors, Alzheimer’s may be predicting an artificial intelligence system developed by researchers in the United States.
The system analyzes brain imaging and early diagnosis is considered critical, as any treatment interventions are more effective if they begin before brain degeneration reaches an advanced stage. But to date, early diagnosis has proven to be very difficult for doctors.
An early predictive method is to detect changes in metabolism due to the disease. But these changes in glucose levels in some areas of the brain are very difficult to detect early. The new “smart” system does exactly this: it can better “read” people about the indiscriminate changes in brain metabolism in Alzheimer’s patients.
The University of California physicians and engineers who published the Radiology journal of the North American Radiological Society trained a deep learning algorithm to analyze positron emission tomography (PET) scanning. The algorithm was then tested in 40 patients and was 100% successful in early detection of Alzheimer’s, on average six years before the final diagnosis.
The researchers said they were satisfied with the accuracy of the system but pointed out that the patient sample was small and the algorithm had to be tested in a larger number of people. If its credibility is confirmed, it will be a valuable assistant to radiologists and other doctors in the future, along with other diagnostic biochemical and imaging examinations.
In addition to detecting changes in glucose, researchers will train the artificial intelligence system to “read” the brain early changes in β-amyloid and tau levels, indicative of Alzheimer’s disease.