Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Parallel qMRI Reconstruction from 4x Accelerated Acquisitions

Mingi Kang

TL;DR

This work addresses the challenge of speeding up MRI by enabling 4\times accelerated parallel imaging through an end-to-end deep learning framework that jointly estimates coil sensitivity maps and reconstructs images from undersampled k-space without precomputed sensitivities. The proposed two-module architecture combines a Coil Sensitivity Map estimator with an MRI reconstruction module, optimized by a composite loss that enforces both image-domain fidelity and k-space consistency. On simulated 4\times data, the method achieves performance close to 2x SENSE (within ~1 dB PSNR) while producing visibly smoother images than SENSE; however, real 4\times acquisitions reveal a sim-to-real gap driven by spatial misalignment and k-space distribution shifts. The work highlights practical deployment challenges and proposes concrete directions—such as registration-based alignment, domain adaptation, and perceptual losses—that could bridge the gap and enable clinically robust, calibration-free accelerated MRI.

Abstract

Magnetic Resonance Imaging (MRI) acquisitions require extensive scan times, limiting patient throughput and increasing susceptibility to motion artifacts. Accelerated parallel MRI techniques reduce acquisition time by undersampling k-space data, but require robust reconstruction methods to recover high-quality images. Traditional approaches like SENSE require both undersampled k-space data and pre-computed coil sensitivity maps. We propose an end-to-end deep learning framework that jointly estimates coil sensitivity maps and reconstructs images from only undersampled k-space measurements at 4x acceleration. Our two-module architecture consists of a Coil Sensitivity Map (CSM) estimation module and a U-Net-based MRI reconstruction module. We evaluate our method on multi-coil brain MRI data from 10 subjects with 8 echoes each, using 2x SENSE reconstructions as ground truth. Our approach produces visually smoother reconstructions compared to conventional SENSE output, achieving comparable visual quality despite lower PSNR/SSIM metrics. We identify key challenges including spatial misalignment between different acceleration factors and propose future directions for improved reconstruction quality.

Parallel qMRI Reconstruction from 4x Accelerated Acquisitions

TL;DR

This work addresses the challenge of speeding up MRI by enabling 4\times accelerated parallel imaging through an end-to-end deep learning framework that jointly estimates coil sensitivity maps and reconstructs images from undersampled k-space without precomputed sensitivities. The proposed two-module architecture combines a Coil Sensitivity Map estimator with an MRI reconstruction module, optimized by a composite loss that enforces both image-domain fidelity and k-space consistency. On simulated 4\times data, the method achieves performance close to 2x SENSE (within ~1 dB PSNR) while producing visibly smoother images than SENSE; however, real 4\times acquisitions reveal a sim-to-real gap driven by spatial misalignment and k-space distribution shifts. The work highlights practical deployment challenges and proposes concrete directions—such as registration-based alignment, domain adaptation, and perceptual losses—that could bridge the gap and enable clinically robust, calibration-free accelerated MRI.

Abstract

Magnetic Resonance Imaging (MRI) acquisitions require extensive scan times, limiting patient throughput and increasing susceptibility to motion artifacts. Accelerated parallel MRI techniques reduce acquisition time by undersampling k-space data, but require robust reconstruction methods to recover high-quality images. Traditional approaches like SENSE require both undersampled k-space data and pre-computed coil sensitivity maps. We propose an end-to-end deep learning framework that jointly estimates coil sensitivity maps and reconstructs images from only undersampled k-space measurements at 4x acceleration. Our two-module architecture consists of a Coil Sensitivity Map (CSM) estimation module and a U-Net-based MRI reconstruction module. We evaluate our method on multi-coil brain MRI data from 10 subjects with 8 echoes each, using 2x SENSE reconstructions as ground truth. Our approach produces visually smoother reconstructions compared to conventional SENSE output, achieving comparable visual quality despite lower PSNR/SSIM metrics. We identify key challenges including spatial misalignment between different acceleration factors and propose future directions for improved reconstruction quality.

Paper Structure

This paper contains 29 sections, 3 equations, 1 figure, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Traditional Parallel MRI Reconstruction
  4. Deep Learning for MRI Reconstruction
  5. Learning-Based Parallel Imaging
  6. Methodology
  7. Problem Formulation
  8. Network Architecture
  9. Coil Sensitivity Map Estimation Module
  10. MRI Reconstruction Module
  11. Loss Function
  12. Training Details
  13. Data Generation
  14. Optimization
  15. Data Split
  16. ...and 14 more sections

Figures (1)

  • Figure 1: Representative reconstructions from Subject S012, Echo 1, Slices 110 and 120. Normalized for visual output. Metrics shown as PSNR (dB) / SSIM.