Purpose

Magnetic resonance (MR) imaging of large field of views at high spatial resolution results in long acquisition times due to the required matrix size and prolonged sampling time to achieve a sufficient SNR. This is particularly challenging in applications like abdominal imaging, using Cartesian k-space sampling, as the acquisition needs to be completed within a breath-hold to avoid ghosting artefacts due to subject motion. Hence, the acquisition time is limited by the breath hold capability of the patient. Alternatively a radial acquisition scheme may be used. This has more benign motion artefact behavior compared to Cartesian sampling, thus acquisition may be performed during free-breathing and being not limited by not scan time limited. However subject motion during radial acquisition will induce blurring artefacts which are particularly problematic at high resolutions. In this study we demonstrate that using a golden angle stack-of-stars sampling scheme[1], combined with individual reconstruction of multiple motion states and subsequent morphing and combining in image space enables sub millimeter in plane resolution.

Methods

A 3D stack of stars acquisition was used, with Cartesian phase encoding in slice (FH) direction and in-plane radial sampling. A single-echo fat suppressed e-THRIVE sequence was used with TR = 4.6 ms, TE = 1.8 ms, and 10° flip angle. In vivo abdominal images were obtained on a Philips Ingenia 3T scanner with a 12 channel posterior and a 16 channel anterior coil from healthy volunteers, with a FOV of 450 × 450 × 202 mm³ and 0.9 × 0.9 × 2 mm³ voxel size. Golden angle radial sampling scheme was used where the angle increment between two successive readouts is given by 111.25°. 1000 profiles per slice were acquired from free breathing subjects amounting to 3 minutes total scan time. Images from individual coils were separately gridded and combined via an optimal Roemer reconstruction taking the receive coil sensitivities into account. The acquired radial profiles are sorted into four motion states, using a k0 auto-navigator[2], [3]. Profiles from each motion state were reconstructed individually. A modified non-rigid, non-parametric registration method consisting of elastic registration steps was applied to generate motion-corrected images[4], by deforming the individual motion states onto one reference state. The deformed images are then combined onto and averaged with the reference image for increased SNR.

Results

Figure 1 displays the reconstruction results of the individual motion states in axial slice as well as coronal reformat. The translation and deformation of the organs due to respiration is clearly observable. The end inspiration as well as the end expiration state each contain more than 300 profiles, whereas the two other states have less than 200 profiles, resulting in streaking artefacts in the axial slices. The streaking manifests also in the coronal slice, where they appear noise-like. The result of elastic registration, morphing and combination is displayed in the right hand panel of Fig. 2. The blurring is greatly reduced, revealing finer details of the organ structure and increased image sharpness compared to the reconstructed image using all profiles (left in Fig. 2, magnification shown in Fig. 3).

Discussion

The golden angle stack-of-stars acquisition combined with auto-navigated multi-bin reconstruction and subsequent image deformation enables imaging resolution not achievable with Cartesian trajectories. This would allow detection of even very small lesions, while at the same time increasing the patient comfort. The presented method may be further refined, by separating the profiles into more than four motion states or even including cardiac motion into the binning process.

Acknowledgements

No acknowledgement found.