Synopsis

Keywords: Pulse Sequence Design, Lung, UTE, Heart Failure, Cartesian

The Cartesian UTE (CUTE) sequence provides high quality three-dimensional imaging of the lung parenchyma (TE = 140 us) without the requirement for k-space gridding in a short breath-hold. Phantom validation and comparison to an existing UTE sequence (Spiral VIBE) is provided in 64 healthy subjects and in clinical examples.

INTRODUCTION

Quantitative imaging of the lung parenchyma requires ultrashort echo time (UTE) acquisitions, typically with TE ≤ 150 µs to avoid excessive signal decay (T2* ~2ms at 1.5T and ~0.7ms at 3.0T)¹, which necessitates non-Cartesian center-out k-space trajectories. The required Cartesian gridding of k-space prior to image reconstruction is computationally demanding and sensitive to gradient imperfections and timing errors that can give rise to artifacts and signal non-uniformity that may limit or preclude quantitative applications, in particular, quantitative imaging of lung water density.² As an alternative, we propose a three-dimensional (3D) gradient-echo Cartesian UTE (CUTE) acquisition approach that combines two separate half-k-space readouts to enable UTE with k-space sampling directly on a Cartesian grid. The CUTE approach was compared to the available Spiral VIBE UTE³ method in phantoms, in a large cohort of healthy subjects and clinical examples.

METHODS

Figure 1 displays the proposed pulse sequence for the 3D gradient-echo CUTE approach. A non-selective rectangular radiofrequency (RF) excitation pulse enabled full-torso spatial excitation and minimal excitation duration. Each k_y and k_z phase-encoding step was acquired separately with both a positive (k_x­+) and a negative readout gradient, k_x- (Fig. 1A). The readout gradient pre-phaser enabled 50%+ΔK_x coverage of k_x-space for both readout directions for sampling of the middle of k-space over a range ±ΔK_x for both readouts (Fig. 1B). The minimum achievable TE was used for each k-space point, TE (k_y,k_z), with the shortest times at the middle of k-space. The repetition time (TR) was held constant. CUTE was compared to Spiral VIBE UTE1 approach for phantom and in-vivo acquisitions. MRI Acquisition: CUTE - 400x500x300 mm FOV (read-encoding superior-inferior), TE=140μs (260μs at edge of k-space), TR=1.7ms, 1595 Hz/pixel, GRAPPA=2, 128x160x64 matrix (3.125x3.125x5mm, reconstructed to 2.08mm isotropic), 2° flip angle, 13 seconds acquisition (breath-hold). 3T PRISMA and 1.5T AERA scanners (Siemens, Erlangen, Germany) using 18 chest/32 back coil array for reception. Spiral VIBE - 550x550x336mm FOV, TE=50 ms, TR=2.46 ms, 945 Hz/pixel, 192x192x96 (2.86x2.86x3.5mm reconstructed to 1.43x1.43x3.5mm) 10 seconds acquisition (only available at 3T). Phantom Experiments: CUTE and Spiral VIBE acquisitions were performed on a composite water and fat phantom with repeated studies using a 0Hz and 200Hz off-resonance field shift to evaluate off-resonance performance (3T). In-Vivo Experiments: CUTE and Spiral VIBE acquisitions were performed on 65 healthy subjects (end-expiration breath-hold) (3T). One healthy subject was imaged with the CUTE method at 1.5T and 3.0T with additional illustrative clinical cases also at 1.5T. In-Vivo Image Processing: A previously reported processing pipeline was used to perform image normalization to units of relative lung water density (rLWD) and segmentation of the lung parenchyma to provide lung water density quantification.² In addition to average rLWD, total lung parenchyma water volume was calculated as rLWV = ΣrLWD·voxel_volume. Finally, segmented rLWD images for all 64 subjects were resampled on a grid (32(RL)x32(HF)x10(CB) pixels) to generate average regional rLWD maps. Airway signal intensities in the trachea, ideally a noise region, were measured to assess the artifact level, in the same units of rLWD.

RESULTS

Phantom studies illustrate the superior performance of CUTE as compared to Spiral VIBE in terms of reduced sensitivity to blurring from off-resonance effects from global field shifts and fat (Fig. 2). Mean whole-lung values of rLWD and rLWV and total lung volume were similar for CUTE and Spiral VIBE (p>0.05) but with significantly higher artifact level (mean airway signal) for VIBE, p<0.05 (Fig. 3). Spiral VIBE had some cases of artifacts and consistent patterns of spatial heterogeneity in rLWD as compared to CUTE that were consistent in all 64 subjects (Fig. 3). Repeated CUTE scans in one volunteer at 1.5T and 3.0T showed consistent appearance and similar rLWD and rLWV values measured over the whole lung volume (Fig. 4). Clinical examples illustrated the measurement of increased water density, globally in a case of chronic heart failure, and regionally in cases of pulmonary hypertension and dilated cardiomyopathy (CUTE, 1.5T) (Fig. 4). Average values for lung water content and volumes were similar between CUTE and Spiral VIBE (Fig. 5). All CUTE image reconstruction was performed online (~25 seconds using the Siemens FIRE application).

DISCUSSION

The Cartesian UTE (CUTE) sequence was shown to provide high quality three-dimensional imaging of the lung parenchyma without the requirement for k-space gridding in a short breath-hold (140μs echo time). In comparison to the available Spiral VIBE UTE sequence, CUTE was shown to have more uniform lung water density values over space and significantly less dependence on off-resonance frequencies. Acquisition of k-space on a Cartesian grid enabled fast on-scanner reconstruction and eliminated the signal non-uniformities observed with the non-Cartesian Spiral VIBE sequence. Image quality and quantitative lung water densities were illustrated to be similar at 1.5T and 3.0T (i.e. field-strength and gradient-performance independent) and clinical cases at 1.5T showed globally and regionally elevated lung water content.

CONCLUSION

The Cartesian UTE (CUTE) sequence provides high quality quantitative three-dimensional imaging of the lung parenchyma without the requirement for k-space gridding in a short breath-hold acquisition.

Acknowledgements

No acknowledgement found.