New York City-based Mount Sinai researchers have developed a deep learning model, HeartBEiT, that can diagnose heart attacks better than established methods for electrocardiogram analysis.
HeartBEiT analyzes ECGs and interprets its analysis as language. Researchers trained the new deep learning model on 8.5 million ECGs from 2.1 million patients from four hospitals within the Mount Sinai Health System.
They then compared HeartBEiT's performance to commonly used machine learning algorithms such as convolutional neural networks in the three cardiac diagnostic areas and found that HeartBEiT could perform significantly higher at lower sample sizes.
According to the researchers, HeartBEiT could highlight the region of the ECG responsible for a diagnosis, such as a heart attack, whereas the convolutional neural networks' explanations were considered "vague,'' according to a June 6 press release from Mount Sinai.
The full study was published June 6 in the medical journal NPJ Digital Medicine.