Today, cardiac disease is one of the most promising cause of mortality. The segmentation of the cardiac image is an essential process to generate personalized models of the heart and to quantify the parameters of cardiac performance. One of the important step is to perform segmentation in Left Ventricle (LR) of cardiac using magnetic resonance images (MRI).However, which can used to find important parameters to be mentioned stroke volume, dischargesection, the structure of the left ventricle myocardium. In addition, the segmentation of the left ventricle helps to build personalized cardiac computer models in order to perform digital simulations. Right now, it is observed that no automated segmentation methods related to cardiac images derived accurate performance. In this article, a new hybrid architectures is proposed where T-Net architecture can combined withSeg-Netto reduced network parameters and used for classification of cardiac MRI images .Then, in order to retrieve an approving performance, we use the EBO (Extended Buffalo Optimization) algorithm to solve the cardiac segmentation. Experimental results show that the proposed method successfully segments LR and achieves 90% accuracy in the cardiac images.