Undersampled free breathing cardiac perfusion MRI reconstruction without motion estimation

One of the key MRI methodologies to identify and characterize coronary artery disease is dynamic contrast enhanced myocardial perfusion imaging. Rapid acquisition of images can help in improved diagnosis by accurately measuring temporal dynamics of the injected contrast agent. Another competing requirement is complete coverage of the heart with high spatial resolution to better identify sub-endocardial infarcts and to reduce dark rim artifacts. Most undersampled reconstruction methods that break this spatio-temporal tradeoff are sensitive to inter-frame respiratory motion in dynamic images. Here we extend a Non-local means based reconstruction to obtain high quality images from undersampled data that has motion. The framework does not rely on estimating motion prior to reconstruction. The method is able to resolve ghosting artifacts from Cartesian undersampling and streaking artifacts due to radial undersampling better than existing methods. The method has potential to improve cardiac perfusion imaging for better diagnosis.