Improving the image quality of cardiac T1 maps via a residual dense network

摘要

Introduction: Recently, a novel fast MR imaging sequence was developed to acquire cardiac 3D T1 maps in a single-breath hold - 3D saturation recovery compressed SENSE rapid acquisition (3D SACORA) [1]. Although this sequence is capable of acquiring T1 maps with good image quality, we hypothesize that the cardiac T1 maps can be further improved using artificial intelligence techniques. A new residual dense network (RDN) has been proposed to improve the resolution and quality of real-world images [2, 3]. In this work, we aim to test the RDN on cardiac T1 maps. Here, we evaluate the improvement in image quality of applying the RDN on the 3D T1 maps obtained with 3D SACORA and the potential effect on quantitative T1 values.