Three-dimensional motion corrected sensitivity encoding reconstruction for multi-shot multi-slice MRI: Application to neonatal brain imaging

Abstract

Purpose: To introduce a methodology for the reconstruction of multi-shot, multi-slice magnetic resonance imaging able to cope with both within-plane and through-plane rigid motion and to describe its application in structural brain imaging.

Theory and methods: The method alternates between motion estimation and reconstruction using a common objective function for both. Estimates of three-dimensional motion states for each shot and slice are gradually refined by improving on the fit of current reconstructions to the partial k-space information from multiple coils. Overlapped slices and super-resolution allow recovery of through-plane motion and outlier rejection discards artifacted shots. The method is applied to T₂ and T₁ brain scans acquired in different views.

Results: The procedure has greatly diminished artifacts in a database of 1883 neonatal image volumes, as assessed by image quality metrics and visual inspection. Examples showing the ability to correct for motion and robustness against damaged shots are provided. Combination of motion corrected reconstructions for different views has shown further artifact suppression and resolution recovery.

Conclusion: The proposed method addresses the problem of rigid motion in multi-shot multi-slice anatomical brain scans. Tests on a large collection of potentially corrupted datasets have shown a remarkable image quality improvement. Magn Reson Med 79:1365-1376, 2018. © 2017 The Authors Magnetic Resonance in Medicine published by Wiley Periodicals, Inc. on behalf of International Society for Magnetic Resonance in Medicine. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

Keywords: image reconstruction; magnetic resonance; motion correction; multi-shot multi-slice images; neonatal brain imaging.

Figure 1

Block diagram of the alternating minimization approach with dynamic outlier rejection.

Figure 2

Illustration of the slice and shot sampling structure for both the T ₂ and T ₁ sequences. a: Acquisition order in time. Colorbar shows the order in which the scanned information is acquired, so the images reflect the order in which the phase encode's (horizontal axis) and slices (vertical axis) are acquired during scan time. This acquisition structure is used in our method to define a set of motion states for which corresponding acquired information is assumed to be subject to negligible motion inconsistencies. b: Sampling structure in the phase encode direction. Colorbar shows the phase encode ordering, so the images reflect the order in which the different shots (horizontal axis) and echoes (vertical axis) covered the acquired k‐space. This spectral acquisition structure is used by our method to infer motion estimates from the partial k‐space information corresponding to each shot.

Figure 3

Comparison of results produced in a mildly artifacted T ₂ axial acquisition when different components of the proposed motion corrected reconstruction method are omitted. a: Conventional uncorrected SENSE reconstructions. b: Uncorrected reconstructions when integrating the slice profile filter. c, d: Motion corrected reconstruction excluding one element at a time: (c) no within‐plane motion model; (d) no through‐plane motion model. e: Full motion corrected reconstructions. From top to bottom, axial view, sagittal view, coronal view, and magnified results within the area enclosed in blue in the sagittal view.

Figure 4

Comparison of results produced in a highly artifacted T ₁ axial acquisition when different components of the proposed motion corrected reconstruction method are omitted. a: Uncorrected reconstructions when integrating the slice profile filter. b–d: Motion corrected reconstruction excluding one element at a time: (b) no outlier rejection strategy; (c) no within‐plane motion model; (d) no through‐plane motion model. e: Full motion corrected reconstructions. From top to bottom, axial view, sagittal view, coronal view, and magnified results within the area enclosed in blue in the sagittal view.

Figure 5

Snapshots of T ₂ reconstructions for different subjects, orientations, and locations in the brain. Each column corresponds to a different subject example. a: Uncorrected reconstructions. b: Corrected reconstructions.

Figure 6

Snapshots of T ₁ reconstructions for different subjects, orientations, and locations in the brain. Each column corresponds to a different subject example. a: Uncorrected reconstructions. b: Corrected reconstructions.

Figure 7

Box plots of relative metric change σ, P‐values of a paired right‐tailed sign test, and percentage of cases r in which the metric decreased for the ${\ell }_1$ norm of Db wavelet decompositions and for the GE in motion corrected versus uncorrected reconstructions. Negative values in the paired box plots indicate a decrease in the corresponding metrics when applying motion correction, which has been documented as associated with an improvement in image quality (29,30). a: Axial T ₂. b: Sagittal T ₂.

Figure 8

Box plots of relative metric change σ, P‐values of a paired right‐tailed sign test, and percentage of cases r in which the metric decreased for the ${\ell }_1$ norm of Db wavelet decompositions and for the GE in motion corrected versus uncorrected reconstructions. Negative values in the paired box plots indicate a decrease in the corresponding metrics when applying motion correction, which has been documented as associated with an improvement in image quality (29,30). a: Axial T ₁. b: Sagittal T ₁.

Figure 9

Slice to volume reconstruction‐based assembling of motion corrected information of different views for suppression of residual artifacts and isotropic resolution: T ₁ example. Volumetric data consistency is substantially improved for each of the views after applying our method, but residual motion may still be present due to remaining inconsistencies between slices, and non‐native views may appear blurred as compared to native views. After slice to volume reconstruction correction, information is made consistent between views and a nearly isotropic representation of the imaged volume is obtained. In the bottom right corner, we show a magnified example comparing the results for the skull in the sagittal slice of the axial acquisition (left image, enclosed in blue) versus the corresponding results after view assembling (right image, enclosed in green), with residual motion inconsistencies strongly suppressed in the latter.