Artificial intelligence (AI) can be used to see the early signs of Alzheimer's disease several years before a patient is usually diagnosed.
Scientists conducting small pilot studies train self-learning computer programs to recognize story features in brain scans that are too subtle to be seen by humans. This system was able to detect the onset of Alzheimer's in 40 patients on average more than six years before they were diagnosed before.
British AI expert Professor Noel Sharkey, from Sheffield University, said of the US findings: "This is exactly the type of task studied in looking for high-level patterns in the data. Although the sample size and test set are relatively small, the results are very promising so much larger research will be useful. "
American researchers trained "deep learning algorithms" using more than 2,100 PET (positron emission tomography) scans from 1,002 patients.
In one test, the algorithm was given a set of 40 scans from 40 patients that had never been studied before.
This proved 100% accurate in detecting Alzheimer's disease on average more than six years before the patient's final diagnosis.
The research was published in the latest issue of the journal & # 39; Radiology. & # 39;