Cardiac diffusion tensor imaging is an effective way to depict the fiber structure of the myocardium. A navigator(NAV)-based stimulated-echo (STEAM) method was proposed by Nielles-Vallespin to obtain cDTI in vivo. However, its use of a biofeedback process where the subjects had to adapt their breathing pattern in real-time can hinder its clinical implementation. In this abstract, we optimized the NAV accept/reject algorithm, using which the scanning efficiency and the image SNR were both largely improved. Therefore, our work laid a great foundation for the clinical use of free breathing cDTI in the future.