Artificial intelligence detects Alzheimer's disease six years earlier than doctors

<pre>Artificial intelligence detects Alzheimer's disease six years earlier than doctors

There are many ways of early diagnosis of Alzheimer's disease of varying degrees of accuracy, and one of them is a PET scan of the brain using fluorodeoxyglucose. Initially, the technique was developed for the diagnosis of certain types of cancer, so scientists have learned to analyze in detail the process of absorption of glucose by the brain. And they began to identify patterns in the images, which were later identified as Alzheimer's disease biomarkers.

The problem is that the differences between how a healthy and slightly sick brain absorbs glucose are visually very little different. While there are no obvious deviations, the human eye simply has nothing to cling to in a snapshot of a multitude of thin, diffuse lines. But machine vision is another matter, and together with the algorithms of deep learning, it has allowed British scientists to re-look at the problems of early diagnosis of the disease.

As part of the study, a new neural network received as a textbook 2100 images of the human brain with Alzheimer's signs in its early stages. Having studied, she managed to build a model of development of deviations in the course of metabolic processes from insignificant fragments to large nodes of the disease. And to learn how to predict such a development of events — this allowed us to identify signs of a dangerous disease 6 years before the transformation of small defects into large and obvious symptoms.

(function (d, s, id) {   var js, fjs = d.getElementsByTagName (s) [0];   if (d.getElementById (id)) return;   js = d.createElement (s); = id;   js.src = “//”;   fjs.parentNode.insertBefore (js, fjs); } (document, 'script', 'facebook-jssdk'));

Source link