The computer was “taught” to diagnose Alzheimer’s disease

Scientists were able to create a program for predicting dangerous disease.

Alzheimer’s disease, probably would not be so dangerous, if it can be recognized at a very early stage. However, her clinical symptoms (like problems with memory, etc.) arise when the pathological processes in the brain has gone relatively far.

It would seem that we now know what kind of processes such as the accumulation in the nervous tissue specific protein deposits; and why not try to consider their brain off when they start to appear? Because modern methods like positron emission tomography (PET), is it possible to see a similar protein abnormalities which are accompanied by massive death of neurons.

But the human eye is often difficult to see the results of brain imaging dangerous areas – different people protein deposits formed in different ways, and not always their appearance entails Alzheimer’s disease. On the other hand, even if a person has little cognitive impairment, they do not necessarily indicate it on Alzheimer’s. If the disease still occurs, between the first abnormalities seen in the nerve tissue, and clinical signs may take from five to ten years.

However, if it is impossible to predict Alzheimer’s disease “by eye”, then it might be worth to try your luck with artificial intelligence? Researchers from McGill University have developed an algorithm, which with 84% accuracy allows to fix alzgeimerovsky processes before they can be diagnosed with conventional methods.

The program analyzes the results of the PET, and at first she was trained on the results of the PET scan about two hundred patients who have already manifested a slight cognitive problems. Then the program showed the brain scans of the same patients who did not earlier than two years before they were diagnosed with Alzheimer’s. Comparing the brain before the diagnosis and after, the machine had to learn to see the disease, yet hidden from other diagnostic methods.

Finally, in the last stage of the program showed tomographic scans of the brain of two hundred and seventy people with some cognitive problems – some of them two years later, really was diagnosed with Alzheimer’s. The machine, however, about the final diagnosis did not know, however, she is able to recognize developing the disease in 84% of cases. The results were published in Neurobiology of Aging.

In the future, as writes portal LiveScience, machine prediction of Alzheimer’s disease, should be tested on a larger number of people to gain more powerful stats and at the same time to improve the accuracy of the algorithm. However, according to the authors of the work, with Alzheimer’s in principle impossible to create such a diagnostic method, which will work with absolute precision, and if we want more confidence in diagnosis, we need to use several diagnostic techniques.

Now we have had a number of experimental tools that inhibit alzgeimerovsky processes in the brain, however, their efficiency highly depends on when they began to apply. Obviously, with a new machine algorithm can reliably test the effectiveness of such measures in clinical trials, and at the same time give a chance to many people, if not to overcome the disease, or at least stop it in reasonable limits.

© 2017 – 2019, paradox. All rights reserved.