A lot of medical conditions typically have symptoms leading up to the actual disease, and because of this, sometimes patterns are useful at identifying a potentially more serious issue at hand. This is what researchers have been trying to do when they trained an AI to help diagnose Alzheimer’s years before a doctor would have.
In a study published in the Radiology journal (via Engadget), researchers in California have developed an AI system that when trained, was capable of diagnosing a patient with Alzheimer’s based on brain scans taken years before. The AI was trained using existing patients who have been confirmed to have Alzheimer’s, and what the researchers found was that the AI system could detect the disease much earlier on compared to an actual physician.
This was used with FDG-PET images which has been used in detection in the past, where FDG is a radioactive glucose that is injected in a person’s bloodstream. The glucose is absorbed into the body, and because of its radioactivity, a PET scan can tell a tissue’s metabolic activity based on how much of the FDG has been absorbed, which can also be used as evidence of Alzheimer’s.
That being said, while the initial tests have proven to be successful, the researchers admit that it is still too small an amount of testing data for it to be really conclusive, and that they do plan on using their algorithm on larger data sets in the future.
Filed in . Read more about AI (Artificial Intelligence), Health and Science.