Artificial intelligence can be used to accurately examine electrocardiogram (ECG) test results, according to the findings of two preliminary studies being presented at the American Heart Association Scientific Sessions 2019 in Philadelphia, PA.
In the first study, researchers evaluated 1.1 million ECGs that did indicate atrial fibrillation (AF) from more than 237,000 patients. They used specialized computational hardware to train a deep neutral network to assess 30,000 data points for each respective ECG.
The results showed that approximately one in three people received an AF diagnosis within a year. Moreover, the model demonstrated the capacity for long-term prognostic significance as patients predicted to develop AF after one year had a 45% higher hazard rate in developing AF over a follow-up duration of 25-years compared to other patients.
The second study included an analysis of the results of 1.77 million ECGs and other records from approximately 400,000 patients. The researchers compared data to machine learning-based models that either directly analyzed ECG signals or relied on human-derived measures. According to the results, the machine-learning model exhibited efficacy at predicting one-year risk of death. Importantly, the researchers observed that the neural network was able to accurately predict risk of death in patients who were deemed by physicians to have a normal ECG.
“This is the most important finding of this study,” said Brandon Fornwalt, MD, PhD, senior author on both studies and associate professor and chair of the Department of Imaging Science and Innovation at Geisinger in Danville, Pennsylvania in a press release. “This could completely alter the way we interpret ECGs in the future.”
Dr. Fornwalt added that this study “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.”