Single-shot High-resolution Free-breathing Cardiac Dark-blood Imaging Based on Model-Incorporated Transfer Learning

Motivation: Routine multi-shot T2-weighted dark-blood fast spin echo (T2W DB-FSE) is hindered by susceptibility to respiratory motion and the requirement for breath-holding. Goal(s): To develop a single-shot cardiac T2W-DB imaging technique with a short echo train to eliminate breath-holding. Approach: We designed a single-shot DB-FSE sequence with a deep learning-based reconstruction model. Transfer learning was employed to pre-train the model on large cardiac MR datasets and fine-tune it on clinical DB-FSE data. Validation was performed with single-shot imaging in 20 healthy volunteers. Results: The transfer learning strategy enhanced model performance with limited data. Single-shot T2W DB-FSE showed image quality comparable to conventional multi-shot imaging. Impact: The proposed single-shot imaging overcome the challenge of conventional cardiac T2W DB-FSE. It has the potential for clinical application in the assessment of myocardium edema on patients having trouble with long breath-holding.