2137
Approaches for correcting motion in diffusion-weighted imaging acquired using diffusion gradient cycling
Boyan Xu1, Shaojun Hu2, Yang Fan1, Bing Wu1, and Ming Song2,3,4
1MR Research, GE Healthcare, Beijing, China, 2Brainnetome Center, Institute of Automation, Chinese Academy of Sciences, Beijing, China, 3University of the Chinese Academy of Sciences, Beijing, China, 4Laboratory of Brain Atlas and Brain-inspired Intelligence, Institute of Automation, Chinese Academy of Sciences, Beijing, China
1MR Research, GE Healthcare, Beijing, China, 2Brainnetome Center, Institute of Automation, Chinese Academy of Sciences, Beijing, China, 3University of the Chinese Academy of Sciences, Beijing, China, 4Laboratory of Brain Atlas and Brain-inspired Intelligence, Institute of Automation, Chinese Academy of Sciences, Beijing, China
Synopsis
Keywords: Diffusion Analysis & Visualization, Motion Correction, Diffusion Gradient Cycling
Motivation: Diffusion gradient cycling (DGC) enables more efficient diffusion-weighted imaging (DWI) scanning, but it is not compatible with typical preprocssing pipelines.
Goal(s): Our goal is to propose two approaches for correcting motion in DWI data acquired using DGC: slice-to-volume registration and volume-to-volume registration with the assistance of slice reordering.
Approach: Intentional motion was introduced during the DWI acquisition, and the proposed approaches were implemented and applied to remove artifacts caused by this motion.
Results: Both approaches effectively eliminated motion-induced artifacts, and the intentional motion was estimated correctly.
Impact: Motion-induced artifacts can be eliminated, and correct motion estimation can be achieved in DWI acquired with DGC. Our proposed approaches are publicly available and can be easily integrated into preprocessing pipelines.
Introduction
The need to explore the microstructure of neural tissue and connectional anatomy across various length scales drives the advancement of high-performance gradient technology1–3. The required strong gradients pose significant demands regarding thermal heating and power supply, resulting in longer acquisition times to allow for cooling periods.To address this issue, diffusion gradient cycling (DGC) has been implemented to control the playout order of the slice stack and different diffusion weighting directions so that the required add-on time for gradient cooling is reduced when diffusion directions are interleaved across the series (Figure 1). This hence enables more efficient scanning. However, DGC distributes the slices affected with motion. This slice misalignment results in the corruption in all volumes (Figure 2A). Existing motion correction tools based on volume-to-volume registration are incompatible with DGC4.
In this study, two approaches were proposed and evaluated for correcting motion in diffusion-weighted imaging (DWI) acquired using DGC.
Methods
Two approaches for motion correction are illustrated in Figure 2. Slice-to-volume-based registration methods have been implemented to correct artifacts caused by within-volume motion-induced artifacts4,5, which is also expected to correct the slice misalignment introduced by DGC. An alternative approach involves reordering all slices before and after volume-to-volume registration to prevent slice misalignment during the registration precedure.One participant underwent scanning on a 3T GE SIGNA UHP scanner (GE Healthcare, Waukesha, Wisconsin, US) for two runs using a single-shot spin-echo echo planar imaging sequence with the following parameters: TE/TR=47.7ms/10s, acquisition matrix=120×120, 60 slices, 2mm isotropic voxel size, 1 image with b-value=0s/mm2 and 30 noncolinear directions with b-value=1000s/mm2. Automated DGC was disabled in the first run but enabled in the second run. During each run, the subject was instructed to rapidly move to different positions.
Before motion correction, image denoising (MRtrix’s dwidenoise) and removing Gibbs ringing artefacts (MRtrix’s mrdegibbs) were applied. Volume-to-volume-based correction and movement estimation were performed using FSL's eddy6. Slice-to-volume registration was performed using both FSL’s eddy and the SHARD reconstruction software7,8. The slice reordering was implemented in python and can be found at GitHub (https://github.com/xu-boyan/gradient_cycling_reorder). Finally, DTI parameter maps were calculated to evaluate the effectiveness of preprocessing.
Results
In the case of DGC, motion-induced positional differences were distributed across slices within each volume, resulting in the misalignment between slices within the same volume (Figure 3). The estimation of intentional motion was erroneous due to this slice misalignment. However, with our proposed slice reordering method, correct estimation of intentional motion was achieved (Figure 4). After applying the regular preprocessing pipeline to data acquired with DGC, motion-induced artifacts were evident in the calculated parameter maps. These artifacts can be eliminated with our suggested motion correction approaches (Figure 5).Discussion
Two approaches for correcting motion in DWI acquired using DGC have been proposed and evaluated in this study. Both effectively eliminated motion-induced artifacts compared to the typical volume-to-volume-based method. Slice-to-volume-based registration methods have been implemented in various software packages, and our source code for slice reordering is publicly available. Therefore, these approaches can be easily integrated into preprocessing pipelines. Estimating subject motion correctly is beneficial for quality control in large imaging datasets.It should be noted that further calculation may also be affected since the b-matrix cannot be properly rotated when correcting for subject motion in data acquired with DGC9. Slice-wise calculation is therefore necessary to achieve a high degree of precision.
Conclusion
Motion-induced artifacts can be eliminated, and correct motion estimation can be achieved in DWI acquired with DGC. Our proposed approaches are publicly available and can be easily integrated into preprocessing pipelines.Acknowledgements
No acknowledgement found.References
1. Foo TKF, Tan ET, Vermilyea ME, et al. Highly Efficient Head-Only Magnetic Field Insert Gradient Coil for Achieving Simultaneous High Gradient Amplitude and Slew Rate at 3.0T (MAGNUS) for Brain Microstructure Imaging. Magn. Reson. Med. 2020;83:2356–2369.2. Huang SY, Witzel T, Keil B, et al. Connectome 2.0: Developing the next-Generation Ultra-High Gradient Strength Human MRI Scanner for Bridging Studies of the Micro-, Meso- and Macro-Connectome. NeuroImage 2021;243:118530.
3. Fan Q, Eichner C, Afzali M, et al. Mapping the Human Connectome Using Diffusion MRI at 300 mT/m Gradient Strength: Methodological Advances and Scientific Impact. NeuroImage 2022;254:118958.
4. Tax CMW, Bastiani M, Veraart J, Garyfallidis E, Okan Irfanoglu M. What’s New and What’s next in Diffusion MRI Preprocessing. NeuroImage 2022;249:118830.
5. Ferrante E, Paragios N. Slice-to-Volume Medical Image Registration: A Survey. Med. Image Anal. 2017;39:101–123.
6. Andersson JLR, Sotiropoulos SN. An Integrated Approach to Correction for Off-Resonance Effects and Subject Movement in Diffusion MR Imaging. NeuroImage 2016;125:1063–1078.
7. Andersson JLR, Graham MS, Drobnjak I, Zhang H, Filippini N, Bastiani M. Towards a Comprehensive Framework for Movement and Distortion Correction of Diffusion MR Images: Within Volume Movement. NeuroImage 2017;152:450–466.
8. Christiaens D, Cordero-Grande L, Pietsch M, et al. Scattered Slice SHARD Reconstruction for Motion Correction in Multi-Shell Diffusion MRI. NeuroImage 2021;225:117437.
9. Leemans A, Jones DK. The B-Matrix Must Be Rotated When Correcting for Subject Motion in DTI Data. Magn. Reson. Med. 2009;61:1336–1349.
Figures
Figure 1. Schematic illustration of the acquisition schemes. (A) The conventional acquisition follows one-volume-one-encoding paradigm, where all slices with the same diffusion encoding are acquired within each TR. (B) Diffusion gradient cycling refers to the variation of diffusion encoding from slice to slice within each TR. After the full acquisition, all slices will be reorganized into volumes based on their diffusion encoding. This illustration disregards the slice acquisition order within one TR.
Figure 2. Two approaches for correcting motion in diffusion-weighted imaging (DWI) acquired using diffusion gradient cycling (DGC). DGC distributes the slices affected with motion, which results in the corruption in all volumes (A). This type of motion induced artifacts is expected to be corrected (C) using either slice-to-volume registration (A-C), or volume-to-volume registration with the assistance of slice reordering. (A-B-D-C).
Figure 3. The effect of gross subject movement. (A) Conventionaly, movement that occurs between volumes leads to positional differences, while the slices are aligned within each volume. (B) In the case of diffusion gradient cycling, motion-induced positional differences are distributed across slices within each volume, and the slices within the same volume are not aligned.
Figure 4. Correct estimation of intentional motion was achieved using FSL's eddy when diffusion gradient cycling (DGC) was disabled (A). When DGC was enabled, the identical motion estimation procedure produced erroneous results (B). With the implementation of our proposed slice reordering method, the motion was estimated correctly (C).
Figure 5. DWI data acquired using the coventional paradigm can be effectively corrected, revealing fine structures in the parameter maps (cyan arrow) (A). When the regular preprocessing pipeline was applied to data acquired with diffusion gradient cycling, motion-induced artifacts remained evident in the parameter maps (yellow arrow) (B). These artifacts can be eliminated with our suggested motion correction approaches (C, D, E).
DOI: https://doi.org/10.58530/2024/2137