New Study Demonstrates AI-Assisted Detection of Reduced Ejection Fraction

Heart failure is often diagnosed at an advanced stage, typically in acute settings when symptoms have already significantly progressed. Since its symptoms can be subtle and non-specific, many cases go undetected until considerable deterioration occurs. A common marker for identifying heart failure with reduced ejection fraction (HFrEF) is ejection fraction (EF), which measures the heart’s ability to pump blood effectively. Now, a new study has explored the potential of an artificial intelligence (AI) platform to help detect reduced EF, a critical indicator of heart failure.

The peer-reviewed study, published in JACC Advances, evaluated the effectiveness of Eko Health’s (San Francisco, CA, USA) FDA-cleared AI model in analyzing heart sounds and single-lead electrocardiogram (ECG) data obtained using a digital stethoscope to identify patients with significantly reduced EF (EF≤40%). The study involved 2,960 adults from four U.S. healthcare networks who were undergoing echocardiography. Data was captured using Eko's ECG-enabled digital stethoscope, ensuring that echocardiograms were performed within one week of data collection. The AI model was then compared with echocardiographic EF measurements, categorizing patients into two groups: normal/mildly reduced EF (>40%) and moderate/severely reduced EF (≤40%). The AI model demonstrated strong predictive capability, with an AUROC of 0.85, and achieved sensitivity and specificity of 77.5% and 78.3%, respectively.

Among the patients flagged by the AI as having potentially low EF but showing EF >40% on their echocardiograms, 25% had EF between 41-49%, and 63% exhibited conduction or rhythm abnormalities, suggesting that the AI model could also play a role in identifying patients who remain at cardiovascular risk. The performance of the AI model was consistent across various demographic and clinical groups, highlighting its broad applicability in clinical settings. The AI model is especially valuable for patients presenting with non-specific symptoms, such as unexplained dyspnea, as it can expedite access to diagnostic testing and treatment. Effective therapies for HFrEF exist and have been shown to improve patient outcomes when initiated early. By using AI-powered heart sound and ECG analysis, clinicians may gain further insights to support timely referrals to specialists, more comprehensive diagnostic evaluations, and better management of the disease.

"The study findings highlight the promise of Eko's platform to complement traditional diagnostics and address the critical challenge of underdiagnosed heart failure," said Connor Landgraf, CEO of Eko Health. "By integrating AI-driven insights into routine physical exams, we can help clinicians identify at-risk patients sooner, particularly in primary care and resource-limited settings."

Related Links:
Eko Health