Myocardial ischemia is intricately linked to adverse cardiovascular outcomes, but current imaging techniques lack objectivity in the diagnosis of myocardial ischemia. Two-dimensional echocardiography (2DE), a widely employed clinical tool, faces challenges in effectively diagnosing the myocardial ischemia due to the issues like unclear myocardial boundaries, intricate tissue structures, and motion artifacts in images. In this study, we present a novel deep neural network model leveraging myocardial bands and its radial motion patterns for the myocardial ischemia diagnosis on left ventricle short-axis (LVSA) 2DE images. Initially, deep learning networks transform the ventricle circle area into a rectangular region more accurately described. Subsequently, the extracted rectangular region, termed the myocardium band, enhances the network’s classification efficiency by capturing myocardial features in radial orientation, considering the myocardium’s radial motion. Finally, the myocardial ischemia is identified using a vision transformer applied to the myocardium bands. Experimental results showcase that our model significantly surpasses original 2DE-based diagnosis, achieving higher accuracy, precision, recall, F1 score, and area under the receiver operating characteristic curve (76.60%, 79.07%, 71.22%, 70.73%, and 0.7994, respectively) compared to the original 2DE results (70.20%, 78.47%, 63.21%, 63.13%, and 0.7011, respectively). This model presents a promising avenue for the diagnosis of the myocardial ischemia, enabling the early detection and treatment of ischemic myocardial diseases.