AI and non-invasive MRI can diagnose Alzheimer's disease long before doctors can

A pair of researchers from the University of Bari in Italy called Nicola Amoroso and Marianna La Rocca and a team of colleagues have developed a new machine-learning algorithm that is able to help diagnose Alzheimer's disease up to nine years before doctors can diagnose the condition. The system the researchers have developed is also able to discern structural changes in the brain that are caused by Alzheimer's.

The team first trained their machine-learning algorithm using 67 MRI scans with 38 of that number being people who suffer from the disease and 29 scans being of healthy control individuals. The scientists want to teach the algorithm to classify and tell the difference between a healthy brain and a brain with Alzheimer's disease. In teaching the algorithm, the team simply divided each of the brain scans into smaller regions and analyzed the neuronal connectivity between the areas with no assumptions made about what an ideal size for these areas might be.

According to the team the most accurate diagnosis for Alzheimer's brains was made when the regions the algorithm studied were in the area of 2250 to 3200 cubic millimeters. La Rocca says that this is a size similar to anatomical structures in the human brain that are associated with Alzheimer's disease. Those structures include the amygdala and hippocampus.

The researchers were them able to turn the algorithm to a second group of brain scans from 148 test subjects. Out of that number, 52 of the scans were of healthy brains and 48 were scans of brains of people who had mild cognitive impairment, and all of that number were known to have developed Alzheimer's disease 2.5 to nine years later.

The AI was able to tell the difference between a brain that was healthy and a brain that has Alzheimer's 86% of the time. More importantly, the AI was able to accurately diagnose brains with mild cognitive impairment, which leads to Alzheimer's, 84% of the time. This means that the AI was able to diagnose Alzheimer's up to a decade before the disease actually manifested. The scientists believe that their AI might be a less costly and more accurate diagnostic tool than currently available tests for Alzheimer's and could possibly be used to diagnose other brain conditions like Parkinson's Disease.

SOURCE: New Scientist