4D MR imaging

Motion imaging and correction is an important task in MRI. We developed a fast and flexible high-resolution 4D (3D + time) MR imaging method of periodic motion, as respiratory or cardiac motion. Since it is not feasible to fill 3D k-spaces of several motion states in finite time under free-movement conditions, we utilize the concept of Compressed Sensing to acquire random samples over time accompanied by a retrospective gating based on a self-navigation signal which yields a 4D subsampled k-space. The gating parameters can all be adjusted retrospectively, providing high flexibility. Furthermore, different surrogate signals can be used in the gating.

Overview

The method aims to provide a clinical feasible setup and is therefore set up in the Gadgetron [Hansen and Sørensen, MRM 2013] framework.

The overall acquistion and reconstruction pipeline is illustrated in the following Figure:

Acquisition

A Cartesian 3D random Gaussian ESPReSSo subsampling is proposed which acquires a random combination of k-space lines ky and kz in each repetition, as illustrated in the Figure below. Periodically the central k-space lines are acquired which serve as a self-navigation signal in the reconstruction.

Reconstruction

After navigator signal extraction from the self-navigation data, the samples are sorted in a gating procedure into their respective motion states. The resulting subsampled k-space is reconstructed via a Compressed Sensing reconstruction.

Download

Source codes

Acquisition

The acquisition sequence, termed CS_Retro (Siemens, VB20P), can be downloaded here: https://github.com/thomaskuestner/4DMRImaging/tree/master/acquisition

The incorporated subsampling is based on: https://github.com/thomaskuestner/CS_LAB/tree/master/acquisition/subsampling_class

Reconstruction

Matlab/Gadgetron: https://github.com/thomaskuestner/4DMRImaging/tree/master/reconstruction

Please cite one of the papers, if you use it in a scientific publication.

References