[ad_1]
One lzheimer occurs in tens of millions of people around the world. Chances are you know someone who has it. The biggest challenge for scientists, however, is the early recognition of the disease. But artificial intelligence can contribute to this.
California scientists have formed a neural network that detects Alzheimer's disease in a small number of patients, reports Engadget. The neural network was able to recognize the disease by means of brain tests that made the actual diagnosis for years.
Special images of brains were used to form the network. These are FDG images, in which a radioactive form of glucose is injected into the blood. The body then incorporates this in the form of ordinary glucose. Scientists then use a scanner to get an idea of the metabolic activity of the tissues, depending on the amount of FDG included.
These scans are already used in the diagnosis of Alzheimer's disease. Affected patients generally have lower metabolic activity in some parts of the brain. The disadvantage is that these images must be analyzed by specialists to recognize the disease, which is difficult because mild cognitive impairments give comparable results.
Results
The team used 2,109 FDG images of 1,002 patients to form the neural network. . For this, they used 90% for training and 10% to test the network. The network was also tested on a separate group of 40 patients. The results of the AI were compared to those of a group of specialists who analyzed the same data.
In the separate data set, the network was able to accurately diagnose the situation in all cases. In 82% of cases, this was also true when a patient did not have the disease. In addition, he predicted diagnoses about six years earlier than when the diagnosis was actually received.
However, scientists claim that the algorithm must first be tested on a much larger dataset before it can be used. But for the future, this promises a lot: the early diagnosis of Alzheimer's disease. As a result, researchers may find better ways to slow down or stop the disease process.
[ad_2]
Source link
