NEJM AI. 2025 May;2(5):AIcs2401008. doi: 10.1056/AIcs2401008. Epub 2025 Apr 24.
ABSTRACT
AbstractAssessment of left ventricular ejection fraction (LVEF) is crucial for prognosis and management in acute coronary syndrome (ACS). While echocardiography is the standard assessment method, assessment via left ventriculography during angiography is often done and carries procedural risks. The CathEF algorithm uses deep learning to estimate LVEF directly from routine left coronary angiograms without additional contrast injection or procedural risks. The CathEF study is a prospective, multicenter, observational study of 207 patients from two Canadian institutions undergoing coronary angiography for ACS. The diagnostic performance of CathEF was evaluated using the area under the receiver operating characteristic curve (AUROC), sensitivity, and specificity; metrics were compared with echocardiography and ventriculography. The AUROC for the CathEF algorithm was 0.84 (95% confidence interval [CI], 0.79 to 0.89) in identifying LVEF less than or equal to 50%, and 0.90 (95% CI, 0.86 to 0.96) in identifying LVEF less than or equal to 40%. Performance was reduced in patients with ST-elevation myocardial infarction (STEMI) (AUROC, 0.76; 95% CI, 0.66 to 0.88) and those with right coronary artery involvement (AUROC, 0.73; 95% CI, 0.60 to 0.86). In summary, the CathEF algorithm offers a reliable, noninvasive tool for real-time screening for LV systolic dysfunction during coronary angiography in patients with ACS. (Funded by the Canadian Institute for Advanced Research Solution Network on Integrated AI for Health Imaging and others.).
PMID:42932536 | DOI:10.1056/AIcs2401008