4339

SENSE-based Multi-shot DWI Reconstruction with Extra-navigated Rigid Motion and Contrast Correction for Brain EPI

Malte Steinhoff¹, Alfred Mertins¹, and Peter Börnert^2,3
¹Institute for Signal Processing, University of Luebeck, Luebeck, Germany, ²Philips Research Europe, Hamburg, Germany, ³Department of Radiology, LUMC, Leiden, Netherlands

Synopsis

We propose an extra-navigated SENSE-based multi-shot DWI reconstruction algorithm that comprises navigator-based phase and rigid in-plane motion corrections at fast reconstruction times. Furthermore, this approach exploits the low-resolution navigator signal to perform diffusion contrast corrections explicitly within the model. The extra-navigated method is compared in-vivo to a self-navigated reference algorithm. The extra-navigated motion estimation from low-resolution navigator data yields decent reconstructions which perfectly coincide with self-navigated results. Moreover, extra-navigation allows for fast reconstruction at the cost of lower scan efficiency and appears to be more robust for strong motion corruption and high segmentations.

Introduction

Multi-shot DWI techniques offer promising potential to improve current state-of-the-art diffusion-weighted imaging (DWI) and diffusion tensor imaging (DTI) applications¹ in terms of SNR and spatial resolution. However, multi-shot DWI is prone to shot-specific variations from tiny physiological motion² as well as inter-shot gross motion. Supported by parallel imaging techniques^3,4, self-navigated^5,6 and navigator-free⁷ reconstruction algorithms have been proposed to overcome the shot-specific variations. Besides those data-driven approaches, extra-navigated methods provide robust alternatives at the cost of decreased scan efficiency, sampling a 2D low-resolution navigator image after a second refocusing pulse⁸. Navigator-based multi-shot DWI reconstructions effectively capture shot phase variations⁸ and rigid in-plane motion⁹ including contrast corrections to account for the changed effective diffusion direction after rotational motion. This work presents a fast extra-navigated multi-shot DWI reconstruction algorithm based on SENSE and further exploits the navigator signal for low-resolution diffusion contrast corrections.

Theory

The extra-navigated macroscopic motion-corrected algorithm solves a multi-shot DWI problem that differs from a previous approach⁹ by introducing navigator-based contrast corrections into the model and adopting a SENSE⁴-based formulation:
$$ \hat{\mathbf{\rho}} = \underset{\mathbf{\rho}}{\mathrm{argmin}} \quad \frac{1}{2} \sum_{i \in I} \lVert F_i S \Omega_i \Phi_i C_i \; \mathbf{\rho} - \mathbf{d}_i \rVert_2^2 + \alpha \lVert \mathbf{\rho} \rVert_2^2,$$
where $$$F_i$$$, $$$\Omega_i$$$, $$$\Phi_i$$$ and $$$C_i$$$ are the Fourier, macroscopic motion, physiological motion and contrast correction operator of shot $$$i$$$, respectively. $$$\mathbf{\rho}$$$ is the joint image, $$$S$$$ the sensitivity operator, $$$\mathbf{d}_i$$$ the i-th shot data, $$$\alpha$$$ the Tikhonov regularization parameter and $$$I$$$ the set of included shots from $$$N_i$$$ total shots to allow for data rejection. This model is solved for every diffusion direction $$$k$$$ of $$$N_k$$$ total directions.

Algorithms

First, the navigator data is upsampled to full resolution by a triangular window and reconstructed using SENSE⁴. Next, the rigid motion, shot phases and data rejection criteria are subsequently calculated from the navigator images. Then, diffusion tensor estimates are calculated from the navigator magnitudes after affine in-plane alignment to calculate shot-wise low-resolution diffusion contrast correction maps according to the estimated rotation. Note that this DTI analysis is performed using all $$$N_k N_i$$$ navigator shots with their respective rotation-corrected diffusion direction^6,9. As there have not been any shot combinations up to this DTI estimation, unlike the approach by Dong et al⁹, the proposed approach remains unbiased. Finally, the multi-shot model is solved for the joint image estimate $$$\hat{\mathbf{\rho}}$$$ using conjugate gradients (CG). The algorithm is visualized in Fig. 1.

This algorithm is evaluated with respect to IRIS⁸ and a self-navigated algorithm¹⁰. IRIS is an extra-navigated approach that only includes phase operators $$$\Phi_i$$$ into the multi-shot model, whereas the iterative self-navigated approach also corrects for rigid motion $$$\Omega_i$$$. Partial Fourier is included by homodyne reconstruction. Image registration is performed using a normalized gradient field metric¹¹. The shot with the highest correlation to all other shots is chosen as the reference shot for registration and data rejection. Shots below 95% correlation are excluded.

Methods

Multi-shot echo-planar diffusion data was acquired with an extra 2D navigator using a twice-refocused Stejskal-Tanner sequence⁸. The data was obtained from 8 healthy volunteers on a 3T Philips Ingenia using a 13-channel head coil. Informed consent was obtained according to the rules of the institution. DTI data was acquired with {4, 5} shots, b = {0, 1000} s/mm² in 15 diffusion directions with a resolution of 1x1x4 mm³. The DTI experiments were performed twice, first the subjects remained still and second continuous in-plane motion was performed. Fractional anisotropy (FA) maps were calculated using Dipy¹² after affine alignment.

Results

Figure 2 compares the proposed extra-navigated to the self-navigated¹⁰ approach for a motion-corrupted dataset. Figure 3 shows the full sets of 5-shot DWI reconstructions within a DTI dataset for a severely motion-corrupted case. The data was registered to the mean rigid motion direction within each multi-shot group. The multi-shot images thus remain misaligned. A diffusion contrast example is provided in Fig. 4. Figure 5 compares extra-navigated DTI reconstructions under static and dynamic conditions. The extra-navigated method took on average about 15 s and 17 s with contrast correction disabled and enabled, whereas the self-navigated approach needed on average 52 s with 55 iterations.

Discussion

The rigid motion estimates of the self- and the extra-navigated approaches coincide remarkably well (Fig. 2). The low-resolution navigators prove to enable sufficiently accurate macroscopic motion estimation. In extreme cases with severe rigid motion and high segmentation, the extra-navigated approach was found to be more robust due to the reliable low-resolution information and the equality of the navigator trajectories for each shot (Fig. 3). The contrast corrections are in a range around and below 5%. The evaluation of the contrast correction in simulations is subject to future research (Fig. 4). Both approaches could be further extended by subsequent high-resolution diffusion contrast corrections, as previously suggested⁹. The extra-navigated approach recovers important structural brain information in the presence of strong in-plane motion and provides fast reconstructions and robust navigators at the cost of less scan efficiency (Fig. 5).

Conclusion

The proposed SENSE-based extra-navigated multi-shot approach efficiently recovers decent diffusion images and proves to be a robust alternative to self-navigated algorithms, especially for strong in-plane motion and high segmentations paving the way for future clinical use.

Acknowledgements

No acknowledgement found.

References

1. Wu W and Miller KL. Image formation in diffusion MRI: A review of recent technical developments: Review of Image Formation in dMRI. JMRI. 2017;46(3):646–662.

2. Miller KL and Pauly JM. Nonlinear Phase Correction for Navigated Diffusion Imaging. MRM. 2003;50:343-353.

3. Roemer PB, Edelstein WA, Hayes CE, Souza SP, Mueller OM. The NMR phased array. MRM. 1990;16(2):192–225.

4. Pruessmann et al. SENSE: sensitivity encoding for fast MRI. MRM. vol. 1999;42(5):952–962.

5. Guo et al. POCS‐enhanced inherent correction of motion‐induced phase errors (POCS‐ICE) for high‐resolution multishot diffusion MRI. MRM. 2016;75(1):169-180.

6. Guhaniyogi et al. Motion Immune Diffusion Imaging Using Augmented MUSE for High-Resolution Multi-Shot EPI. MRM. 2016;75:639-652.

7. Mani M, Jacob M, Kelley D, Magnotta V. Multi-shot sensitivity-encoded diffusion data recovery using structured low-rank matrix completion (MUSSELS): Annihilating Filter K-Space Formulation for Multi-Shot DWI Recovery. MRM. 2017;78(2):494-507.

8. Jeong H-K, Gore JC, Anderson AW. High-resolution human diffusion tensor imaging using 2-D navigated multishot SENSE EPI at 7 T. MRM. 2013;69(3):793-802.

9. Dong Z, Wang F, Ma X, Dai E, Zhang Z, Guo H. Motion-corrected k-space reconstruction for interleaved EPI diffusion imaging: Motion Correction for iEPI DWI. MRM. 2018;79(4):1992-2002.

10. Steinhoff M, Nehrke K, Mertins A, Börnert P. Multi-shot Diffusion EPI Reconstruction with Iterative Rigid Motion-correction and Motion-induced Phase-correction for Brain Imaging. In: Proceedings of the 27th Joint Annual Meeting of ISMRM. Montréal, QC, Canada; 2019.

11. Kabus S and Lorenz C. Fast elastic image registration. Medical Image Analysis for the Clinic: A Grand Challenge. 2010:81-89.

12. Garyfallidis E, Brett M, Amirbekian B, et al. Dipy, a library for the analysis of diffusion MRI data. Frontiers in neuroinformatics. 2014;8:8.

Figures

Figure 1: Reconstruction scheme for SENSE-based multi-shot DWI with extra-navigated motion and contrast correction. First, the moderately subsampled navigator images are recovered by SENSE. Second, the navigators are used to calibrate shot-wise macroscopic motion, physiological motion and data rejection. Third, navigator diffusion tensors are calculated from all diffusion directions after affine alignment to correct for diffusion contrast changes after rotational shot motion. All these parameters are included into a multi-shot SENSE model to reconstruct the joint image.

Figure 2: Comparison of self- and extra-navigated rigid motion corrections for a 4-shot DTI dataset. a: Colored FA maps for reconstructions without, with extra-navigated and with self-navigated rigid motion estimation. b: Results of the motion estimation approaches. The DTI dataset contains substantial in-plane motion, which is nicely corrected by both methods. The motion parameters are almost identical, confirming that the low-resolution navigator offers sufficiently accurate motion estimates.

Figure 3: Comparison of extra- and self-navigated 5-shot DTI reconstructions for extreme in-plane motion. Extra-navigated (left) and self-navigated (right) multi-shot reconstructions of a complete 15-direction DTI dataset with severe in-plane motion. The diffusion directions are numbered chronologically. The extra-navigated approach is able to robustly recover all datasets while the self-navigated approach contains two corrupted results indicated by orange arrows (9: strong rigid mismatch, 14: blurring).

Figure 4: Diffusion contrast correction example for 4-shot DTI data. Upper row: Colored FA maps for low-resolution navigator and image data. An exemplary shot contrast correction map with strong in-plane rotation is shown aside. Bottom row: Associated multi-shot reconstruction with and without contrast correction. The difference image reveals consistent structures, especially in highly anisotropic brain regions. The accuracy evaluation of this approach in simulations is subject to future research.

Figure 5: Comparison of extra-navigated motion-corrected reconstructions for static and motion-corrupted datasets. a: Colored FA maps for a static (left) and a motion-corrupted case (center and right), reconstructed once without (IRIS⁸) and once with rigid motion correction. b: Rigid motion estimates. Subject 1 performed only minor gross motion throughout the scan (±1 mm, ±2 deg), whereas the second example is heavily motion corrupted (±5 mm, ±10 deg). The extra-navigated rigid motion correction recovers great parts of the structural information visible in the static example.

Proc. Intl. Soc. Mag. Reson. Med. 28 (2020)

4339