Gaining on Alzheimer’s: A.I. to the rescue

As many as one in three elderly Americans develop Alzheimer’s Disease. But, until recently, there’s been no way to accurately predict the illness. That means that doctors couldn’t be sure which treatment to try. Now bioscientists at Italy’s University of Bari Aldo Moro have used artificial intelligence to pick out telltale hints of Alzheimer’s in brain scans. The researchers trained the artificial-intelligence program by showing it brain scans of 67 brains, 38 of which were confirmed as having Alzheimer’s. After understanding the difference between a healthy brain and an afflicted one, the program spotted Alzheimer’s brains correctly in 86 percent of cases. The AI program focused on two Alzheimer’s characteristics: the build-up of “sticky” protein plaques and tangles of nerve strands, both of which prevent brain cells from communicating with each other. But what if Alzheimer’s could not only be diagnosed early, but also halted? A Washington University team of scientists may have taken a big step toward that goal. Alzheimer’s is characterized by two proteins in the brain. One, called “tau,” stabilizes nerve-cell structure in healthy brains. In Alzheimer’s, however, tau proteins tangle. The other protein, called B-amyloid, clots and forms plaque in Alzheimer’s brains that confuses messages among brain cells. The two proteins conspire to do the lethal damage that Alzheimer’s is known for, with tau thought to trigger the disastrous effects of B-amyloid. Because people with B-amyloid brain plaque don’t always show signs of the disease, the research team focused on tau – and found a mutant protein called ApoE4, present in 10 percent to 15 percent of people, heightened the dangerous actions of tau. Working with mice, scientists found that reducing the amount of ApoE proteins in the brain, especially ApoE4, significantly reduced chances of developing Alzheimer’s symptoms. Although the finding is too new to have led to possible treatment protocols, it suggests a promising direction for more research.