Physics-Informed Self-supervised Deep Learning Reconstruction for Accelerated First-Pass Perfusion Cardiac MRI

First-pass perfusion cardiac magnetic resonance (FPP-CMR) is becoming an essential non-invasive imaging method for detecting deficits of myocardial blood flow, allowing the assessment coronary heart disease. Nevertheless, acquisitions suffer from relatively low spatial resolution and limited coverage. Compressed sensing (CS) methods have been proposed to accelerate FPP-CMR achieve higher resolution. However, long reconstruction times widespread clinical use CS in FPP-CMR. Deep learning techniques based on supervised emerged as alternatives speeding up reconstructions. these approaches require fully sampled data training, which not possible obtain, particularly high-resolution images. Here, we propose a physics-informed self-supervised deep approach accelerating scans hence facilitate high imaging. The provides high-quality images 10x undersampled without using reference data.