Deep Learning for Assessment of Left Ventricular Ejection Fraction from Echocardiographic Images

For automated measurements of left ventricular ejection fraction (LVEF), obtaining accurate border detection is a difficult task due to the complicated temporal deformation of the left ventricle (LV). Recently, deep learning (DL) has been developed as a state-of-the-art method for the classification of cardiovascular diseases.1,2 Our study aim was to evaluate whether a three-dimensional convolutional neural network (3DCNN) could estimate and differentiate preserved ejection fraction (EF) or reduced EF independently of volumes using echocardiographic images.