Automatic Detection of Cardiac Ejection Rate in Left Ventricular Photography Using Deep Learning

Left ventricular angiography is an assessment of cardiac structure and function that is often performed during cardiac catheterization. In addition to obtaining the index of left ventricular ejection rate (LVEF), left ventricular angiography also assists in the interpretation of cardiac hypertrophy, valve regurgitation, and regional myocardial systolic dysfunction. This also assists the clinician in the treatment and the judgment of opening the blood vessels. Although left ventricular angiography has been a well-established and internationally recognized detection method, it is not easy to train cardiac catheterization professionals, as image analysis requires sufficient experience but it is time-consuming. In this study, the deep learning models are introduced to reduce clinical interpretation errors and provide information on cardiac function and local cardiac systolic abnormalities in a short period. Then, clinicians can easily make medical decisions and judgments. The models in this study are expected to reduce 90% of operation time and improve the accuracy from 50% to 99%. It also helps detect myocardial hypoxia and determine whether to perform coronary surgery. Automatic and accurate detection of cardiac ejection rate, cardiac hypertrophy1, and abnormal local contraction is enabled to immediately warn and feedback to clinicians to make the most accurate treatment decisions.