AI could assist early, rapid diagnosis of asymptomatic heart failure
Key takeaways:
- An AI model may help to detect asymptomatic heart failure using a single-lead ECG.
- Early diagnosis could improve patient outcomes and quality of life.
- Results were consistent across most patient subgroups.
AI-assisted diagnosis of asymptomatic left ventricular dysfunction may detect disease early and thereby expedite treatment of low ejection fraction, according to data published in JACC: Advances.
“Early detection of left ventricular dysfunction is crucial, as delayed diagnosis often leads to worse patient outcomes and higher health care costs,” Salima Qamruddin, MD, technical director of the echocardiography laboratory at the Ochsner Medical Center and director of the Women’s Cardiovascular Clinic at Ochsner Heart & Vascular Institute, said in a press release. “This study demonstrates how AI-enhanced digital stethoscope technology may serve as a powerful tool in identifying patients with potential heart failure earlier, enabling clinicians to take proactive steps in patient management.”
As Healio previously reported, the AI algorithm was designed for HF screening (Eko DUO, Eko) and received breakthrough device designation in 2019 and a newer iteration (Eko Low EF AI) received FDA clearance in April 2024 for the detection of low EF. The algorithm is designed to be used alongside the company’s digital stethoscope (Eko) and detect low EF in approximately 15 seconds.
For the present study, Qamruddin and colleagues tested the algorithm’s ability to detect asymptomatic LV systolic dysfunction in a cohort of 2,960 adults undergoing single-lead ECG and echocardiography at Jefferson Einstein Philadelphia Hospital, Prairie Cardiovascular Consultants, MedStar Health Research Institute and Ochsner Heart & Vascular Institute (mean age, 66 years; 51% women; 25% Black). Patients with a prior diagnosis of HF or reduced EF were not excluded from the study.
The algorithm’s performance in detecting asymptomatic LV dysfunction was compared with the echocardiogram machine’s integrated cardiac quantification software or manual biplane Simpson’s measurement, both of which were subsequently validated by board-certified cardiologists, according to the study.
With an area under the receiver operating characteristic curve of 0.85, sensitivity of 77.5%, specificity of 78.3%, positive predictive value of 20.3% and negative predictive value of 98%, the AI model demonstrated “potential value as a noninvasive, scalable and relatively low-cost method for identifying those at increased risk of reduced EF,” the researchers wrote.
The results were consistent across most subgroups, with some variability observed between men and women, people older than 70 years and those with conduction abnormalities, according to the study.
Moreover, 25% of false-positive screenings for reduced EF had moderately reduced EF — between 41% and 49% — and 63% of false positives had conduction and/or rhythm abnormalities.
“Earlier diagnosis of a previously unrecognized low EF would allow for the earlier initiation of therapies proven to improve quality of life, decrease hospitalizations and prolong survival,” the researchers wrote. “While our findings highlight the promise of this technology for improving detection rates, further research is needed to evaluate its impact on clinical outcomes and cost-effectiveness. Future work will focus on adapting the model for 3-lead input and optimizing algorithm performance using transformer models.”
Reference:
- Eko health announces new study demonstrating AI-assisted detection of reduced ejection fraction. https://www.prnewswire.com/news-releases/eko-health-announces-new-study-demonstrating-ai-assisted-detection-of-reduced-ejection-fraction-302389211.html. Published March 3, 2025. Accessed March 5, 2025.
Collapse