AI-Based Algorithm Enables Quicker HF Diagnosis

A novel deep learning (DL) computer algorithm can identify subtle electrocardiogram (ECG) shifts that predict heart failure (HF), according to a new study.

The DL algorithm, developed at the Icahn School of Medicine at Mount Sinai (MSSM; New York, NY, USA), is designed to quantify left ventricular (LV) and right- ventricular (RV) dysfunction from ECG data in order to assist diagnostic workflow. To do so, a computer read 147,636 patient reports paired to 715,890 ECGs; the written reports acted as a standard set of clinical data so that the computer could compare them with identified patterns in the ECG images, helping the DL algorithm learn how to recognize heart pumping strengths via LV ejection fraction (LVEF).


The results showed the algorithm was 94% accurate at predicting which patients had a healthy LVEF and 87% accurate at predicting those who had an LV ejection fraction that was below 40%. The AI algorithm, however, was only 73% accurate at predicting HF in patients with an LVEF between 40% and 50%. The results also suggested that it was 84% accurate in detecting RV weaknesses, defined by more descriptive terms extracted from the ECG reports. The study was published on October 13, 2021, in JACC: Cardiovascular Imaging.

“We showed that deep-learning algorithms can recognize blood pumping problems on both sides of the heart from ECG waveform data. Ordinarily, diagnosing these type of heart conditions requires expensive and time-consuming procedures,” said senior author Benjamin Glicksberg, PhD, of the Hasso Plattner Institute for Digital Health at MSSM. “We are in the process of carefully designing prospective trials to test out its effectiveness in a more real-world setting. We hope that this algorithm will enable quicker diagnosis of heart failure.”

Clinicians currently rely on ECGs as a diagnostic tool, analyzing data points with the naked eye in order to identify patterns and abnormalities that are characteristic of heart disease. The problem is this takes time, and is entirely reliant on training and expertise. An experienced clinician can be reasonably reliable, but expertise levels vary significantly. Currently, AI tools to estimate cardiac function are restricted to quantification of very low LVEF values, when clinical HF is already evident.

Related Links:
Icahn School of Medicine at Mount Sinai