AI program predicts atrial fibrillation
Today, our system is 97% accurate and identifies anywhere from 1% up to a high-risk group of patients.
New artificial intelligence algorithms can predict people at risk of serious health issues. The detection rate is, expectedly, better than cardiologists. More detectable for the algorithms are atrial fibrillation and premature death. Over the last couple years, this program has seen improvement. Implementation is visible on more robust models so as to identify more accurately who is at-risk. A recent study found that patients with atrial fibrillation may have an increased risk of heart attack or stroke.
A team of researchers at the Geisinger Medical Center in Pennsylvania made two presentations on the subject at the American Heart Association’s annual meeting in Philadelphia, used more than two million electrocardiogram results on some 637,000 patients over the past three decades to train their system, which is based on deep learning neural networks.
The system identified 1% of high-risk patients, including three out of every five who ended up diagnosed with atrial fibrillation in the next year. The algorithm was able to accurately predict the risk of premature death within the same time frame, even for patients whose doctor had judged their electrocardiogram to be normal.
Three doctors then assessed some electrocardiograms that an other doctor had previously judged normal. They couldn’t see what the artificial intelligence had, but again they weren’t able to spot any future risk markers.
When it comes to atrial fibrillation, there are limited possibilities to identify who will develop it within the next year, and more patients are dying of it due to not knowing they’re at risk.
If you enjoyed this article on the AI program, then you may also like this one.
Author: PC-GR
The World of Technology
