Improved In-plane SENSE PROPELLER with Multi-Step Joint-Blade Reconstruction
Mengye Lyu1,2, Yilong Liu1,2, Victor B. Xie1,2, Yanqiu Feng1,2, Zhe Zhang3, Hua Guo3, and Ed X. Wu1

1Laboratory of Biomedical Imaging and Signal Processing, The University of Hong Kong, Hong Kong SAR, China, People's Republic of, 2Department of Electrical and Electronic Engineering, The University of Hong Kong, Hong Kong SAR, China, People's Republic of, 3Department of Biomedical Engineering, Tsinghua University, Beijing, China, People's Republic of

Synopsis

This work addresses three problems for in-plane SENSE-accelerated PROPELLER: generally amplified noise, increased artifact with narrow blades, and degraded motion correction with wide blades. We show that a novel reconstruction method - multi-step join-blade (MJB) SENSE can lead to higher SNR and less narrow-blade artifact than conventional SENSE reconstruction. In addition, k-space truncation is used to improve the motion correction robustness. We also show that additional iterations can enhance MJB SENSE. This new method can greatly benefit future SENSE PROPELLER studies with improved image quality.

Introduction

In PROPELLER, parallel imaging is used mainly in two ways: widening the blades with fixed echo train length (ETL), or shortening the ETL with fixed blade width. Wider blades can reduce scan time and potentially benefit motion correction [1, 2], while shorter ETL [3, 4] can alleviate T2 decay, reduce SAR, and facilitate T1-w contrast. On the other hand, parallel imaging can enhance noise and introduce artifacts. Here, we identify and address three major problems in in-plane SENSE-accelerated PROPELLER: (i) general noise amplification problem, (ii) ringing artifact with narrow blades, and (iii) degraded motion correction with wide blades. A novel reconstruction method, multi-step join-blade (MJB) SENSE, is proposed to reduce noise amplification and narrow-blade artifact. In MJB SENSE, k-space truncation is used to improve motion correction robustness. We also show that additional iterations can enhance MJB SENSE.

Methods

MJB SENSE: Conventionally in SENSE PROPELLER [1], blades are separately reconstructed. We denote this method as simple single-blade (SSB) SENSE. MJB SENSE starts with SSB SENSE, and consists of three steps as follows (Figure 1).

(1) SSB SENSE is performed. It provides rough estimations of blade images, blade-combined images, and motion parameters. To improve robustness, motion estimation [5] is performed after truncating the k-space of blades in phase encoding directions, such that SENSE artifact is smoothed and less harmful.

(2) Regularized single-blade SENSE is performed. It improves the estimations provided by SSB SENSE. Using the estimated blade-combined images, motion parameters, and the phases of blade images, we regenerate two sets of blade images, with and without coil sensitivity map (CSM) weighting. Dividing the CSM-weighted regenerated blade images by non-CSM-weighted ones, we obtain narrow-blade CSMs. Each pixel is re-estimated with “back-substitution”, which means when estimating one pixel, all other pixels are treated known and directly replaced using the regenerated data.

(3) Joint-blade SENSE is performed. By patching regenerated high-frequency data, all acquired blades are widened to have full resolution. Then the blades are registered according to motion parameters. Thus, for each pixel, all the single-blade SENSE equations can be combined into a joint-blade SENSE equation. Back-substitution strategy is also used.

Iterations can be performed by feeding back the output of step 3 to step 2 to update the estimation of blade-combined images.

Simulations: PROPELLER data were synthesized from multi-channel phantom images, which were acquired with matrix size 256-by-256 at 3T using eight-channel coil. Inter-blade in-plane rigid motion was added with uniform random distribution (±5 degrees rotation and ±5 pixels translation). Random Gaussian noise was added with SNR level controlled.

In Vivo Experiments: Human brain data were acquired with fully sampled PROPELLER (3T Philips scanner, eight-channel coil, TR/TE=4000/111ms, matrix size=436×436, 23 blades×30 lines), on which retrospective undersampling was performed. The CSMs were computed using ESPIRIT [6] from the central 24×24 k-space of the fully sampled data.

Results

(1) Motion correction results in wide-blade simulation (Figure 2) showed that conventional motion correction was heavily degraded at SENSE factor 6. Motion correction was largely improved after truncating the blade width to 30.

(2) Noise-free narrow-blade simulation showed (Figure 3) that, SSB SENSE resulted in ringing artifact; MJB SENSE did not suffer from such artifact.

(3) Applied to real brain data (Figure 4), MJB SENSE resulted in obviously less noise than SSB SENSE. With one more iteration, the image quality was further improved.

Discussion and Conclusion

(1) MJB SENSE reduces noise in in-plane accelerated PROPELLER. MJB SENSE forms joint-blade equations and utilizes blade-combined images such that the blades are jointly reconstructed with mutual regularization. Before, with SSB SENSE, more blades are needed to compensate for SNR loss [1]; now with MJB SENSE, fewer blades are needed so that PROPELLER can be truly accelerated.

(2) MJB SENSE also successfully reduces narrow-blade artifact, because the CSMs are customized for narrow blades in step 2, and the blades are widened in step 3. Reducing narrow-blade artifact holds great significance for in-plane acceleration, because shorter ETL and more flexible contrast are directly facilitated.

(3) SENSE artifact can degrade motion correction. Robust motion correction is crucial to the proposed MJB SENSE method, because motion parameters are used to regenerate data. Our results showed that by truncating the k-space data, robust motion correction can be achieved, leading to successful imaging reconstruction using MJB SENSE.

(4) MJB SENSE is actually applicable to both in-plane accelerated and slice accelerated PROPELLER. Reducing noise is more difficult with in-plane acceleration, because only 2D coil sensitivity can be utilized. In this regard, we use additional iterations to enhance MJB SENSE.

Acknowledgements

No acknowledgement found.

References

[1] Chang, Y., et al., Magn Reson Med, 2014.

[2] Skare, S., et al., Magn Reson Med, 2008.

[3] Chuang, T.C., et al., Magn Reson Med, 2006.

[4] Holmes, J.H., et al., Magn Reson Med, 2012.

[5] Pipe, J.G., et al., Magn Reson Med, 2014.

[6] Uecker, M., et al., Magn Reson Med, 2014.

Figures

Figure 1. Procedure of MJB SENSE for in-plane accelerated PROPELLER. Step 2 uses results of step1 for back-substitution (BS) regularization and narrow-blade coil sensitivity maps (NB-CSMs). Step 3 - joint-blade (JS) SENSE can feed back results to step 2 to form iterations. Ib_comb: blade_combined images; motion pars: estimated motion parameters.

Figure 2. Comparison of the motion correction (step 1 results) in wide-blade phantom simulation. Motion correction was better when the motion parameter were estimated from blades that were truncated to 30 lines. Blade number = 16, blade width = 120 (ETL=20), SENSE factor = 6; Noise was added with SNR=10.

Figure 3. Reconstructed images and error maps in noise-free narrow-blade simulation using SSB SENSE and MJB SENSE. SSB led to ringing artifact, which MJB SENSE was mostly free from. Blade number = 16, blade width = 30 (ETL=5), SENSE factor = 6; error maps are displayed × 5.

Figure 4. Reconstructed images and error maps on real human brain ROPELLER data. Compared to SSB SENSE, MJB SENSE largely reduced the noise amplification. One additional iteration further improved image quality. The subject was asked to be still. SENSE factor 6 was used. Error maps are displayed × 2.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
3230