Cardiac MR T1 Mapping with Shortened Acquisition Time and Window via Deep Learning

Motivation: MOLLI is a widely used technique for cardiac MR (CMR) T1 mapping, which, however suffers from long breath-hold and acquisition window. We aim to propose a deep learning network to mitigate these challenges. Goal(s): To develop a deep learning model to simultaneously reduce the acquisition time and window of MOLLI. Approach: A novel transformer model is designed to explore the spatial and inter-contrast correlations the MOLLI T1-weighted images to subsequently generate high-quality CMR T1 maps using only four low-resolution single-shot T1-weighted images. Results: The proposed deep learning model is able to achieve accurate quantification and super-resolution of CMR T1 maps simultaneously. Impact: The proposed method can generate high-quality cardiac T1 maps using MOLLI images acquired with reduced acquisition time (~4s) and three to four-fold shortened acquisition window, thereby enhancing the clinical applicability of cardiac T1 mapping.