Abstract:TARGET AUDIENCE: Researchers and clinicians interested in high resolution DWI/DTI PURPOSE: To improve the spatial resolution of diffusion weighted imaging (DWI), many multi-shot acquisition and reconstruction schemes have been proposed [1]. The shot-to-shot phase variations are acquired or estimated to eliminate ghost artifacts. Multiplexed sensitivity-encoding (MUSE) [1] is a multi-shot EPI based DWI method using SENSE [2] for self-navigating. In MUSE, SENSE is used for solving the phase of each excitation, so the reduction factor (R) equals to the number of shot. When many shots are used to decrease the EPI distortion and blurring, the high R will cause inaccurate estimation of phase by SENSE, and hence introduce artifacts in DW images. In this work, self-feeding MUSE is proposed. By using self-feeding [3] mechanism, robustness of phase estimation and MUSE reconstruction can be improved when higher shot number is used. METHODS: In MUSE for multi-shot EPI DWI reconstruction, SENSE is firstly performed to get an unaliased image for each excitation, and the phase as navigator information is extracted after smoothing. The accuracy of the phase is critical to the ghost elimination and final image quality, so the number of shot, which equals the SENSE R, is limited due to the noise and artifact amplification of high R. In self-feeding MUSE, MUSE is firstly performed (step 1). The magnitude of the MUSE reconstruction and the smoothed phase are combined as the prior image for another SENSE reconstruction using the prior information regularized SENSE scheme [4] (step 2): argmin where is the desired image which contains navigation phase, is the encoding matrix, is the sampled data, is the prior image and is the regularization parameter. Then these improved navigation phases are used for the final MUSE reconstruction (step 3). Fig. 1 shows the schematic diagram of self-feeding MUSE. To demonstrate the advantages of self-feeding MUSE, 6-shot DW EPI data sets with NSA=3 were acquired on a Philips 3T Achieva scanner with an 8-channel head coil. 15 slices were acquired using FOV=220 220mm, voxel size=0.89 0.89 4mm, 6 diffusion directions with b=800s/mm. The initial phase of SENSE reconstruction and improved phase of regularized SENSE were shown in Fig. 2. Reconstructed DW images and color-coded FA maps from original MUSE, self-feeding MUSE and single-shot EPI (2 2. 5 4mm, SENSE R=2) were compared in Fig. 3. RESULTS AND DISCUSSION: Since the regularization term will greatly improve the condition of high R (R=6 with 8-ch coil) SENSE, the phase estimation of regularized SENSE has much lower noise/artifact level (Fig. 2), especially at image center where geometry factor is high. The accurate phase estimation will benefit the final MUSE reconstruction. Fig. 3 compares the reconstructed DW images and FA maps of original MUSE, self-feeding MUSE and the single-shot EPI. Compared to original MUSE, self-feeding scheme reduces the ghost artifact (pointed by blue arrow). For color-coded FA map, self-feeding MUSE has more precise directionality estimation than original MUSE, by comparing with single-shot EPI reference (pointed by yellow arrow). Using the initial MUSE reconstruction results as prior information for regularized SENSE, the phase variations among shots can be calculated more precisely, thus eliminating the ghosts and benefitting the final multi-shot EPI reconstruction. CONCLUSION: Self-feeding MUSE provides robustness and accuracy in phase variance estimation and shows its potential for high resolution DWI without need for extra navigator acquisition. REFERENCES: [1] Bammer, R., et al. Magn Reson Imaging Clin N Am 17(2009), 175. [2] Chen, N., et al. NeuroImage 72 (2013): 41. [3] Pruessmann, K. P., et al. MRM 42.5 (1999): 952. [4] Huang, F., et al. MRM 64.4 (2010): 1078. [5] Lin, F. H., et al. MRM 51.3(2004): 559. Fig 1. Schematic diagram of self-feeding MUSE.

Deep learning based multiplexed sensitivity-encoding (DL-MUSE) for high-resolution multi-shot DWI

Improving the image quality of DWI in breast cancer: comparison of multi-shot DWI using multiplexed sensitivity encoding to conventional single-shot echo-planar imaging DWI

Self-feeding MUSE : A Method for High Resolution Diffusion Weighted Imaging with Robust Phase Variation Estimation

Self-feeding MUSE: A Robust Method for High Resolution Diffusion Imaging Using Interleaved EPI

Simultaneoussuperresolutionreconstruction and Distortion Correction for Single‐shotepi DWIusing Deep Learning

Simultaneous Super-Resolution and Distortion Correction for Single-shot EPI DWI using Deep Learning

Highly Accelerated Multi-Shot Intravoxel Incoherent Motion Diffusion-Weighted Imaging in Brain Enabled by Parametric POCS-based Multiplexed Sensitivity Encoding

<scp>EPI</scp> phase error correction with deep learning ( <scp>PEC‐DL</scp> ) at <scp>7 T</scp>

A Paired Phase and Magnitude Reconstruction for Advanced Diffusion-Weighted Imaging

EPI Phase Error Correction with Deep Learning (PEC-DL) at 7 T.

Distortion correction of single-shot EPI enabled by deep-learning

An artificial intelligence-accelerated 2-minute multi-shot echo planar imaging protocol for comprehensive high-quality clinical brain imaging

Multishot Cartesian Turbo Spin-Echo Diffusion Imaging Using Iterative POCSMUSE Reconstruction.

Multiple B-Value Model-Based Residual Network (MORN) for Accelerated High-Resolution Diffusion-Weighted Imaging

Convolutional network denoising for acceleration of multi-shot diffusion MRI

Bi-Directional MS Lesion Filling and Synthesis Using Denoising Diffusion Implicit Model-based Lesion Repainting

DWI using navigated interleaved multishot EPI with realigned GRAPPA reconstruction

Improved Multi-Shot Diffusion-Weighted MRI with Zero-Shot Self-Supervised Learning Reconstruction

DrDisco: Deep Registration for Distortion Correction of Diffusion MRI with single phase-encoding

Deep Learning with Diffusion Basis Spectrum Imaging for Classification of Multiple Sclerosis Lesions

Simultaneous multislice accelerated interleaved EPI DWI using generalized blipped-CAIPI acquisition and 3D K-space reconstruction.