Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

FINDER: Zero-Shot Field-Integrated Network for Distortion-free EPI Reconstruction in Diffusion MRI

Namgyu Han, Seong Dae Yun, Chaeeun Lim, Sunghyun Seok, Sunju Kim, Yoonhwan Kim, Yohan Jun, Tae Hyung Kim, Berkin Bilgic, Jaejin Cho

Abstract

Echo-planar imaging (EPI) remains the cornerstone of diffusion MRI, but it is prone to severe geometric distortions due to its rapid sampling scheme that renders the sequence highly sensitive to $B_{0}$ field inhomogeneities. While deep learning has helped improve MRI reconstruction, integrating robust geometric distortion correction into a self-supervised framework remains an unmet need. To address this, we present FINDER (Field-Integrated Network for Distortion-free EPI Reconstruction), a novel zero-shot, scan-specific framework that reformulates reconstruction as a joint optimization of the underlying image and the $B_{0}$ field map. Specifically, we employ a physics-guided unrolled network that integrates dual-domain denoisers and virtual coil extensions to enforce robust data consistency. This is coupled with an Implicit Neural Representation (INR) conditioned on spatial coordinates and latent image features to model the off-resonance field as a continuous, differentiable function. Employing an alternating minimization strategy, FINDER synergistically updates the reconstruction network and the field map, effectively disentangling susceptibility-induced geometric distortions from anatomical structures. Experimental results demonstrate that FINDER achieves superior geometric fidelity and image quality compared to state-of-the-art baselines, offering a robust solution for high-quality diffusion imaging.

FINDER: Zero-Shot Field-Integrated Network for Distortion-free EPI Reconstruction in Diffusion MRI

Abstract

Echo-planar imaging (EPI) remains the cornerstone of diffusion MRI, but it is prone to severe geometric distortions due to its rapid sampling scheme that renders the sequence highly sensitive to field inhomogeneities. While deep learning has helped improve MRI reconstruction, integrating robust geometric distortion correction into a self-supervised framework remains an unmet need. To address this, we present FINDER (Field-Integrated Network for Distortion-free EPI Reconstruction), a novel zero-shot, scan-specific framework that reformulates reconstruction as a joint optimization of the underlying image and the field map. Specifically, we employ a physics-guided unrolled network that integrates dual-domain denoisers and virtual coil extensions to enforce robust data consistency. This is coupled with an Implicit Neural Representation (INR) conditioned on spatial coordinates and latent image features to model the off-resonance field as a continuous, differentiable function. Employing an alternating minimization strategy, FINDER synergistically updates the reconstruction network and the field map, effectively disentangling susceptibility-induced geometric distortions from anatomical structures. Experimental results demonstrate that FINDER achieves superior geometric fidelity and image quality compared to state-of-the-art baselines, offering a robust solution for high-quality diffusion imaging.

Paper Structure

This paper contains 14 sections, 2 equations, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Method
  3. Problem Formulation of Joint Reconstruction
  4. Network Architecture
  5. Physics-Informed INR Field Estimator.
  6. Dual-Domain Unrolled Reconstruction Network.
  7. Zero-Shot Optimization Strategy
  8. Experiments and Results
  9. Experimental Setup
  10. High-Fidelity Distortion-Free Reconstruction
  11. Diffusion Analysis and Field Refinement
  12. Discussion and Conclusion

Figures (4)

  • Figure 1: Overview of the proposed FINDER framework. The upper part presents the INR network for field map updates, while the lower part details the physics-guided unrolled image reconstruction network.
  • Figure 2: Visual comparison of the diffusion-weighted images at 5-fold acceleration (single direction). Yellow insets highlight baseline artifacts effectively suppressed by FINDER.
  • Figure 3: Visual comparison of the average DWIs across 32 directions and the corresponding FA maps. Yellow insets highlight severe noise amplification and artifacts in baseline methods, which are effectively suppressed by FINDER.
  • Figure 4: Visual comparison of the estimated field maps. TOPUP and FD-net estimate the field maps based on the highly aliased initial SENSE reconstructions. In contrast, FINDER dynamically refines the initial TOPUP field map using the proposed INR network. Red arrows indicate severe estimation errors, which are effectively mitigated by FINDER. The RMSEs of field maps in the brain region are 19.43 Hz, 31.07 Hz, and 13.57 Hz for TOPUP, FD-net, and FINDER, respectively.