Coronary heart disease is the primary cause of death among adults globally. The standard diagnostic procedure used in clinical decision-making, ranging from medication prescriptions to coronary bypass surgery, is coronary angiography. Therefore, assessing the left ventricular ejection fraction (LVEF) during coronary angiography is crucial for optimizing treatment decisions, especially in cases of acute coronary syndromes (ACS) that pose a potential threat to life.
CathEF pinpoints left ventricular ejection fraction during procedures
The study published in JAMA Cardiology examined the potential of using deep neural networks (DNNs), an AI algorithm, to predict cardiac pump function from standard angiogram videos. Led by Dr Geoff Tison and Dr Robert Avram, the researchers developed and tested a DNN called CathEF. CathEF aims to estimate the left ventricular ejection fraction (LVEF) using coronary angiograms of the left side of the heart. This could provide a non-invasive alternative to the current invasive procedure of left ventriculography, which carries risks and increases contrast exposure.
CathEF is an innovative approach that uses AI to analyze data collected during angiograms, providing clinicians with valuable information not currently available. The method involved training a neural network using a cross-sectional study of 4,042 adult angiograms and corresponding transthoracic echocardiograms. The network accurately predicted left ventricular ejection fraction (LVEF) and demonstrated strong correlations with echocardiographic measurements. The algorithm’s performance was also validated using real-world angiograms from the Ottawa Heart Institute across various patient demographics and clinical conditions.
Scientists introduce a novel way to assess LVEF during angiography
The study has presented a novel method for assessing left ventricular ejection fraction (LVEF) during routine coronary angiography without additional procedures or cost. This method is intended to assist in decision-making and patient care. The algorithm was trained on a large dataset of angiograms from UCSF and validated using data from the Ottawa Heart Institute. Further research is being conducted to test the algorithm in real-time settings and assess its impact on the clinical workflow for heart attack patients. In addition, a multi-centre validation study is currently underway to compare the performance of the new method with traditional techniques.