Artificial Intelligence-Based Algorithm Predicts Heart Failure

Mount Sinai researchers created an artificial intelligence-based computer algorithm that can detect subtle changes in electrocardiograms to predict whether a patient is experiencing heart failure.

“We showed that deep-learning algorithms can recognize blood pumping problems on both sides of the heart from ECG waveform data,” Assistant Professor of Genetics and Genomic Sciences Benjamin S. Glicksberg, PhD, said in a press release.

“Ordinarily, diagnosing these type of heart conditions requires expensive and time-consuming procedures. We hope that this algorithm will enable quicker diagnosis of heart failure.”

Impacting about 6.2 million Americans, heart failure, or congestive heart failure, occurs when the heart pumps less blood than the body typically needs. For years, doctors have relied on echocardiograms to assess whether a patient could be experiencing heart failure. While this method is helpful, it is also labor-intensive and not widely available.

According to researchers, however, recent breakthroughs in artificial intelligence signify that an electrocardiogram could be a quicker and more readily available option. Many studies have indicated that a deep learning algorithm could detect weakness in the heart’s left ventricle.

In this study, the researchers described the development of an algorithm that analyzes not only the left but also the right ventricle.

“Although appealing, traditionally it has been challenging for physicians to use ECGs to diagnose heart failure. This is partly because there is no established diagnostic criteria for these assessments and because some changes in ECG readouts are simply too subtle for the human eye to detect,” said Girish N. Nadkarni, MD, MPH, CPH, associate professor of medicine at the Icahn School of Medicine at Mount Sinai.

“This study represents an exciting step forward in finding information hidden within the ECG data which can lead to better screening and treatment paradigms using a relatively simple and widely available test.”

For this study, the researchers programmed a computer to read patient electrocardiograms and data taken from written reports that summarized the results of corresponding echocardiograms taken from the same patients. The written reports acted as a standard set of data for the computer to compare with electrocardiogram data and learn to identify weaker hearts.

Natural language processing programs assisted the computer in extracting data from written reports. Additionally, special neural networks capable of discovering patterns in images were incorporated to train the algorithm to recognize pumping strengths.

The computer read more than 700,00 electrocardiograms and echocardiogram reports obtained from 150,000 Mount Sinai Health System patients between 2003 and 2020. While data from the four hospitals were used to train the computer, data from a fifth was used to test the algorithm.

“A potential advantage of this study is that it involved one of the largest collections of ECGs from one of the most diverse patient populations in the world,” said Nadkarni.

The initial results indicated that the algorithm effectively predicted which patients would have either healthy or very weak left ventricle. Strength was defined by left ventricle ejection fraction, an estimate of how much fluid the ventricle pumps out with each beat. Health hearts have an ejection fraction of 50 percent or greater while weak hearts are equal to or below 40 percent.

The algorithm was 94 percent accurate in predicting which patients had a health ejection fraction and 87 percent accurate at predicting those with an ejection fraction below 40 percent. However, the algorithm was not as accurate at predicting which patients had slightly weakened hearts at 73-percent accuracy.

The results also suggest that the algorithm learned to detect right valve weakness from the electrocardiograms with 84 percent accuracy.

According to researchers, the additional analysis could advance AI’s ability to detect heart weakness in all patients, regardless of race and gender.

“Our results suggest that this algorithm could be a useful tool for helping clinical practitioners combat heart failure suffered by a variety of patients,” said Glicksberg. “We are in the process of carefully designing prospective trials to test out its effectiveness in a more real-world setting.”