Physics-informed self-supervised deep learning reconstruction for accelerated first-pass perfusion cardiac MRI

TL;DR

This paper introduces a physics-informed self-supervised deep learning method for accelerated first-pass perfusion cardiac MRI, enabling high-quality image reconstruction from highly undersampled data without needing fully sampled references.

Contribution

It presents a novel self-supervised deep learning approach that leverages physics constraints to reconstruct high-resolution cardiac MRI images from highly undersampled data.

Findings

Achieves high-quality images from 10x undersampled data

Does not require fully sampled reference data

Speeds up FPP-CMR reconstruction process

Abstract

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