Although Alzheimer's disease affects tens of millions of people worldwide, it is still difficult to identify at an early stage. However, researchers in the field of artificial intelligence in medicine have discovered that this technology can help to diagnose an early-grave illness. The California team recently published a report on its research in the radiology journal and showed that sometimes nerve networks trained were able to accurately diagnose Alzheimer's disease in a limited number of patients based on imaging imaging that took place years before the patient was actually diagnosed by a doctor.
The team uses brain imaging (FDG-PET imaging) to train and test its neural network. In FDG, the patient's circulatory images are injected with radioactive glucose, and then his body tissues, including the brain, push it on the surface. Scientists and doctors can then use PET scans to understand the metabolism of these tissues, depending on how much FDG is taken.
The FDG-PET method is used to diagnose Alzheimer's disease and patients whose disease usually causes a lower metabolic activity in some of the brain. However, experts need to analyze these pictures to find signs of the disease and it becomes very difficult because moderate cognitive impairment and Alzheimer's disease can produce similar scan results.
Therefore, the team uses 1009 patients from 2 109 FDG-PET images, training 90% of nerve networks and testing their remaining 10%. She also runs tests in one set of 40 patients scanned between 2006 and 2016, then compares the results of artificial intelligence with expert group data analyzing the same data.
With a separate set of test data, Artificial Intelligence is able to diagnose Alzheimer's patients with 100% accuracy and 82% accuracy for those who do not suffer from false diseases. He can also predict on average more than six years in advance. For comparison, a group of doctors who studied the same scanned images identified 57% of patients with Alzheimer's and 91% of patients without a disease. However, differences in machine-to-human performance are less significant when it comes to diagnosing mild cognitive impairment of Alzheimer's disease.
Researchers note that their research has a number of limitations, including a limited number of test data and limited types of training data.