Friday , June 18 2021

AI recognizes the years of Alzheimer's before a diagnosis



Shooting of human brains through PET

Thousands of PET images of Alzheimer's patients used the researchers to train their AI. (Photo: Radiological Society of North America)

BerlinIn the fight against Alzheimer's disease, early detection is particularly important. If still incurable dementia is detected early, you can at least slow down your course with medication.

"If we diagnose Alzheimer's only when clear symptoms appear, the loss of cerebral volume is so great that it is usually too late for effective intervention," explains Jae Ho Sohn.

Along with his team at the University of California in San Francisco, the doctor developed a new tool for the early detection of Alzheimer's disease: an adaptive algorithm that predicts reliability of dementia years before the diagnosis of a doctor.

The researchers focused their development on subtle metabolic changes in the brain that are caused by the onset of the disease. These changes can be visualized using an image technique known as positron emission tomography (PET).

Seven factors that favor Alzheimer's

However, the footprints in the early stages of the disease are so weak that they are barely recognizable even to experienced doctors. "It is easier for humans to find specific biomarkers of the disease," explains Sohn. "But metabolic changes are much more subtle processes."

The researchers trained their artificial intelligence using data from the Neuroimaging Initiative on Alzheimer's Disease (ADNI). Among other things, this data collection contains thousands of PET images of Alzheimer's patients in very early stages of the disease. 90 percent of these engravings, researchers used to train the algorithm, the remaining 10 percent to control success.

For the final test, the AI ​​had to finally analyze 40 images that had not been sent until then. The result describes the child as follows: "The algorithm was able to reliably detect any case, which later reached the onset of Alzheimer's disease."

In addition to the success rate of 100 percent, doctors overwhelmed the initial identification of cases. On average, the system recognized the symptoms more than six years before the actual diagnosis of the disease. "We were delighted with this result," they say. However, the doctor also knows that the test series was still relatively small and the subsequent tests must confirm the result.

Still, he sees in his algorithm the potential of an important tool in the treatment of Alzheimer's: "If we can detect the disease earlier, it will give researchers the opportunity to find better ways to slow down or even stop the process of the disease."


Source link