A Geisinger study finds AI can examine ECG results to identify patients who are at risk of dying within a year or at risk of developing abnormal heart rhythms.
It’s been found recently by Geisinger researchers that artificial intelligence can examine electrocardiogram (ECG) test results to identify patients at risk of dying within a year or developing a potentially dangerous type of arrhythmia, or irregular heartbeat, according to a release from Geisinger.
Conducted in two studies, researchers used more than 2 million ECG test results from archived medical records within the Geisinger system to train deep neural networks and predict irregular heart rhythms, known as atrial fibrillation (AF), before they developed. Atrial fibrillation is associated with an increased risk of heart attack and stroke.
The studies are among the first to use artificial intelligence to predict future events from ECG results rather than to detect current health problems.
“This is exciting and provides more evidence that we are on the verge of a revolution in medicine where computers will be working alongside physicians to improve patient care,” says Brandon Fornwalt, MD, PhD, co-senior author on both studies and associate professor and chair of the Department of Imaging Science and Innovation at Geisinger.
In the first study, researchers worked with patients who had not yet developed AF.
Out of 1.1 million ECGs from more than 237,000 patients, the research team used highly specialized computational hardware to train a deep neural network to analyze 15 segments of data - 30,000 data points - for each ECG. Researchers found within the top 1% of high-risk patients, one out of every three people was diagnosed with AF within a year.
The model predictions also demonstrated longer term prognostic significance, as the patients predicted to develop fibrillation had a 45% higher hazard rate in developing fibrillation over the next 25 years of follow-up compared to the lower risk patients.
“Currently, there are limited methods to identify which patients will develop AF within the next year, which is why, many times, the first sign of AF is a stroke,” says Christopher Haggerty, PhD, assistant professor in the Department of Imaging Science and Innovation at Geisinger and co-senior author on both studies. “We hope that this model can be used to identify patients with atrial fibrillation very early so they can be treated to prevent stroke.”
In the second study, researchers studied to identify patients who are most likely to die of any cause within a year.
Geisinger researchers analyzed results of 1.77 million ECGs and other records from almost 400,000 patients. The team used this data to compare machine learning-based models that either directly analyzed the raw ECG signals or relied on aggregated human-derived measures and commonly diagnosed disease patterns.
The neural network model that directly analyzed the ECG signals was found to be superior for predicting 1-year risk of death. Fortunately, the neural network was able to accurately predict risk of death even in patients deemed by a physician to have a normal ECG. Three cardiologists separately reviewed the ECGs that had first been read as normal, and they were generally unable to recognize the risk patterns that the neural network detected, the release says.
“This is the most important finding of this study,” says Fornwalt, who co-directs Geisinger’s Cardiac Imaging Technology Lab with Haggerty. “This could completely alter the way we interpret ECGs in the future.”