Abstract
Medical imaging plays a crucial role in clinical diagnostics, with modalities like Transthoracic Echocardiography (TTE) and Cardiac Magnetic Resonance (CMR) offering distinct advantages and limitations in cardiovascular assessment. While TTE provides real-time, non-invasive imaging, it is operator-dependent and may yield incomplete views. In contrast, CMR offers comprehensive evaluations but is time-consuming and costly. This paper proposes a novel architecture for synthesizing CMR images from TTE inputs using an integrated autoencoder and vision transformer. The autoencoder captures TTE patterns and transforms them into CMR-like representations, enhanced by the vision transformer's attention mechanisms. Evaluation through quantitative and qualitative metrics demonstrates the system's ability to generate realistic CMR images, potentially enhancing diagnostic accuracy and workflow efficiency in cardiac imaging.