Multi‐shot diffusion‐weighted MRI reconstruction with magnitude‐based spatial‐angular locally low‐rank regularization (SPA‐LLR)

Hu, Yuxin; Wang, Xiaole; Tian, Qiyuan; Yang, Grant; Daniel, Bruce L.; McNab, Jennifer A.; Hargreaves, Brian A.

doi:10.1002/mrm.28025

Cited by 28 publications

(35 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The fact that ghosting varies across diffusion direction suggests that it may be advisable to reconstruct the whole set of dMRI images simultaneously by exploiting similarities of kspace acquired at different diffusion directions or q-space values. Those approaches are becoming increasingly popular in the dMRI literature MRI (Ramos-Llordén, et al, 2020) (Wu, Koopmans, Andersson, & Miller, 2019) (Mani, Magnotta, & Jacob, 2021) (Hu, et al, 2020) (Ramos-Llordén, Aja-Fernández, Liao, Setsompop, & Rathi, 2019) (Haldar, et al, 2013) (Wu, Koopmans, Andersson, & Miller, 2019) (Wu, et al, 2014). Exploratory analyses along these lines are an important aspect of our future work.…”

Section: Discussionmentioning

confidence: 99%

High-fidelity, high-spatial-resolution diffusion MRI of the ex-vivo whole human brain on the 3T Connectom scanner using structured low-rank EPI ghost correction

Ramos‐Llordén

Lobos

Kim

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Diffusion MRI (dMRI) of whole, intact, fixed postmortem human brain at high spatial resolution serves as key bridging technology for 3D mapping of structural connectivity and tissue microstructure at the mesoscopic scale. Ex vivo dMRI offers superior spatial resolution compared to in vivo dMRI but comes with its own technical challenges due to the significantly reduced T2 relaxation times and decreased diffusivity incurred by tissue fixation. The altered physical properties of fixed tissue necessitate the use of alternative acquisition strategies to preserve SNR and achieve sufficient diffusion weighting. Multi-shotor segmented 3D echo planar imaging (EPI) sequences have been used to shorten echo times (TEs) with reduced distortions from field inhomogeneity and eddy currents on small-bore MR scanners and have been adopted for high b-value dMRI of ex vivo whole human brain specimens. The advent of stronger gradients on human MRI scanners has led to improved image quality and a wider range of diffusion-encoding parameters for dMRI but at the cost of more severe eddy currents that result in spatial and temporal variations in the background magnetic field, which cannot be corrected for using standard vendor-provided ghost correction solutions. In this work, we show that conventional ghost correction techniques based on navigators and linear phase correction may be insufficient for EPI sequences using strong diffusion-sensitizing gradients in ex vivo dMRI experiments, resulting in orientationally biased dMRI estimates. This previously unreported problem is a critical roadblock in any effort to leverage scanners with ultra-high gradients for high-precision mapping of human neuroanatomy at the mesoscopic scale. We propose an advanced reconstruction method based on structured low-rank matrix modeling that reduces the ghosting substantially. We show that this method leads to more accurate and reliable dMRI metrics, as exemplified by diffusion tensor imaging and high angular diffusion imaging analyses in distributed neuroanatomical areas of fixed whole human brain specimens. Our findings advocate for the use of advanced reconstruction techniques for recovering unbiased metrics from ex vivo dMRI acquisitions and represent a crucial step toward making full use of strong diffusion-encoding gradients for neuroscientific studies seeking to study brain structure at multiple spatial scales.

show abstract

Section: Discussionmentioning

confidence: 99%

High-fidelity, high-spatial-resolution diffusion MRI of the ex-vivo whole human brain on the 3T Connectom scanner using structured low-rank EPI ghost correction

Ramos‐Llordén

Lobos

Kim

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Our method uses a different architecture compared with previous method MoDL‐MUSSELS, 31 and it enables further acceleration of the reconstruction using only one single gradient update and a deeper network (U‐Net) in each iteration. Moreover, our proposed method offers denoising capabilities similar to averaging multiple repetitions using results from a joint reconstruction 17 as the target. Inspired by previous KIKI‐net, 38 we also introduce the cross‐domain‐CNN concept for multishot DWI reconstruction, in which the input space alternates between k‐space and image space.…”

Section: Discussionmentioning

confidence: 99%

“…For shot‐LLR, 200 iterations and a regularization parameter of 0.008 were used 14 . For SPA‐LLR, 100 iterations, a regularization parameter of 0.05, and other settings as in Hu et al 17 were used. The reconstruction results of the joint reconstruction (SPA‐LLR) were used as the target in training.…”

Section: Methodsmentioning

confidence: 99%

“…Additionally, a nonlinear model with spatial-angular LLR regularization (SPA-LLR) 17 has been proposed for joint reconstruction of all diffusion-encoding directions. This method achieves significantly improved denoising performance compared with the conventional single-direction reconstruction methods such as shot-LLR ( Figure 1C,D).…”

Section: Shot-llr and Spa-llr Reconstructionmentioning

confidence: 99%

See 1 more Smart Citation

RUN‐UP: Accelerated multishot diffusion‐weighted MRI reconstruction using an unrolled network with U‐Net as priors

Tian

et al. 2020

Magnetic Resonance in Med

Self Cite

View full text Add to dashboard Cite

Purpose To accelerate and improve multishot diffusion‐weighted MRI reconstruction using deep learning. Methods An unrolled pipeline containing recurrences of model‐based gradient updates and neural networks was introduced for accelerating multishot DWI reconstruction with shot‐to‐shot phase correction. The network was trained to predict results of jointly reconstructed multidirection data using single‐direction data as input. In vivo brain and breast experiments were performed for evaluation. Results The proposed method achieves a reconstruction time of 0.1 second per image, over 100‐fold faster than a shot locally low‐rank reconstruction. The resultant image quality is comparable to the target from the joint reconstruction with a peak signal‐to‐noise ratio of 35.3 dB, a normalized root‐mean‐square error of 0.0177, and a structural similarity index of 0.944. The proposed method also improves upon the locally low‐rank reconstruction (2.9 dB higher peak signal‐to‐noise ratio, 29% lower normalized root‐mean‐square error, and 0.037 higher structural similarity index). With training data from the brain, this method also generalizes well to breast diffusion‐weighted imaging, and fine‐tuning further reduces aliasing artifacts. Conclusion A proposed data‐driven approach enables almost real‐time reconstruction with improved image quality, which improves the feasibility of multishot DWI in a wide range of clinical and neuroscientific studies.

show abstract

“…31 The presented framework assumes for each diffusion direction that the induced contrast variations are negligible under sufficiently small rotations. Contrast corrections can be introduced by imposing q-space relations 2,31,70 between the shots.…”

Section: Unmodeled Shot Variationsmentioning

confidence: 99%

Segmented simultaneous multi‐slice diffusion‐weighted imaging with navigated 3D rigid motion correction

Steinhoff

Setsompop

Mertins

et al. 2021

Magnetic Resonance in Med

View full text Add to dashboard Cite

Purpose To improve the robustness of diffusion‐weighted imaging (DWI) data acquired with segmented simultaneous multi‐slice (SMS) echo‐planar imaging (EPI) against in‐plane and through‐plane rigid motion. Theory and Methods The proposed algorithm incorporates a 3D rigid motion correction and wavelet denoising into the image reconstruction of segmented SMS‐EPI diffusion data. Low‐resolution navigators are used to estimate shot‐specific diffusion phase corruptions and 3D rigid motion parameters through SMS‐to‐volume registration. The shot‐wise rigid motion and phase parameters are integrated into a SENSE‐based full‐volume reconstruction for each diffusion direction. The algorithm is compared to a navigated SMS reconstruction without gross motion correction in simulations and in vivo studies with four‐fold interleaved 3‐SMS diffusion tensor acquisitions. Results Simulations demonstrate high fidelity was achieved in the SMS‐to‐volume registration, with submillimeter registration errors and improved image reconstruction quality. In vivo experiments validate successful artifact reduction in 3D motion‐compromised in vivo scans with a temporal motion resolution of approximately 0.3 s. Conclusion This work demonstrates the feasibility of retrospective 3D rigid motion correction from shot navigators for segmented SMS DWI.

show abstract

Multi‐shot diffusion‐weighted MRI reconstruction with magnitude‐based spatial‐angular locally low‐rank regularization (SPA‐LLR)

Cited by 28 publications

References 35 publications

High-fidelity, high-spatial-resolution diffusion MRI of the ex-vivo whole human brain on the 3T Connectom scanner using structured low-rank EPI ghost correction

High-fidelity, high-spatial-resolution diffusion MRI of the ex-vivo whole human brain on the 3T Connectom scanner using structured low-rank EPI ghost correction

RUN‐UP: Accelerated multishot diffusion‐weighted MRI reconstruction using an unrolled network with U‐Net as priors

Segmented simultaneous multi‐slice diffusion‐weighted imaging with navigated 3D rigid motion correction

Contact Info

Product

Resources

About