Click here to


Are you sure ?

Yes, do it No, cancel

Model-Based Iterative Reconstruction for Diffusion Weighted Echo Planar Imaging (DW-EPI)

U Yarach*, M Bernstein , J Huston , M In , D Kang , Y Shu , N Meyer , E Gray , J Trzasko , Department of Radiology, Mayo Clinic, Rochester, MN


(Tuesday, 7/16/2019) 3:45 PM - 4:15 PM

Room: Exhibit Hall | Forum 8

Purpose: Diffusion-weighted (DW) magnetic resonance images (MRI) are typically acquired using accelerated echo-planar-imaging (EPI). As standard (e.g., vendor-provided) image reconstruction methods do not prospectively account for B0-inhomogeneities or noise amplification, diffusion-weighted images can exhibit geometric distortion and low signal-to-noise ratio (SNR) – particularly with high-resolution imaging. This work describes a regularized model-based iterative reconstruction (MBIR) framework for accelerated DW-EPI acquisitions that jointly accounts for multiple acquisition non-idealities, and offers improved geometric accuracy and SNR relative to standard reconstruction.

Methods: A healthy volunteer was imaged under IRB approval on a compact 3T [1] (700T/m/s, 80mT/m) MRI scanner with an accelerated high-resolution DWI protocol (single-shot EPI, in-plane=0.86x0.86mm, b={0,1000}, 4x SENSE, 6/8 partial Fourier, NEX=1) and 32-channel phased-array head coil. The T2 and DWI images were individually reconstructed from raw data using the (wavelet) sparsity-regularized MBIR strategy described in [2], which prospectively and simultaneously accounts for ramp sampling, gradient nonlinearity (GNL), and off-resonance effects, and utilizes a computationally efficient type-III NUFFT-based optimization strategy [3]. The regularization parameters for the b=0 and b=1000 images were 0.01 and 0.005, respectively. The B0 field map was generated from a dual-echo calibration scan using graph cuts-based statistical estimation [4].

Results: Our experimental results demonstrate that the proposed MBIR framework -- when appropriately parameterized – effectively mitigates geometric distortion from GNL and off-resonance, and improves SNR without resolution loss. Figure 1 shows that the MBIR result is geometrically is better aligned with a spin-warp reference image than the standard reconstruction result, particularly near air-tissue interfaces. Figure 2 shows that sparsity regularization enables >80% improvements in SNR over the standard reconstruction. The mean diffusivity image calculated from the MBIR results also exhibits noticeably higher SNR compared to that obtained from the standard results.

Conclusion: Sparsity-regularized MBIR simultaneously mitigates geometric distortion and improves image SNR in accelerated single-shot EPI-based DW-MRI.

Funding Support, Disclosures, and Conflict of Interest: This work was supported in part by NIH Grant U01-EB024450


Not Applicable / None Entered.


IM- MRI : Diffusion MRI

Contact Email