 This paper proposes a novel deep learning base tool for automatically segmenting the anatomical structures of the left heart from echocardiogram images. The tool was trained and tested on a data set of 450 patients with echocardiograms from the University Hospital of St. Etienne. The tool achieved high accuracy in segmenting the left heart's atria, ventricles, and endocardial and epicardial layers with dice similarity coefficients ranging from 92.63% to 87.57%. This tool can provide valuable support for cardiologists in their daily practice, allowing them to quickly and accurately assess the health of the left heart. This article was authored by MHD Jaffa Morteda, Selene Tomesini, Haidar Anbar, and others.