B-FIRE: Binning-Free Diffusion Implicit Neural Representation for Hyper-Accelerated Motion-Resolved MRI

2026-01-07

arXiv.org

AI Summary

This work addresses the challenge of accurately capturing instantaneous 3D abdominal dynamics in CINE MRI, which is hindered by artifacts introduced through respiratory phase binning in conventional 4D MRI. The authors propose a diffusion implicit neural representation (Diffusion-INR) framework that eliminates the need for phase binning, leveraging a CNN-based INR encoder–decoder architecture combined with diffusion optimization to enable high-quality reconstruction of dynamic 3D MRI from extremely undersampled data. By integrating a composite loss function that enforces both image-domain fidelity and k-space perceptual constraints, the method achieves state-of-the-art performance in reconstruction quality, motion trajectory consistency, and inference speed. Evaluated on T1-weighted StarVIBE liver MRI, it significantly outperforms existing approaches including NuFFT, GRASP-CS, and unrolled CNNs, marking the first demonstration of bin-free, highly accelerated dynamic MRI.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Representation Learning
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Accelerated dynamic volumetric magnetic resonance imaging (4DMRI) is essential for applications relying on motion resolution. Existing 4DMRI produces acceptable artifacts of averaged breathing phases, which can blur and misrepresent instantaneous dynamic information. Recovery of such information requires a new paradigm to reconstruct extremely undersampled non-Cartesian k-space data. We propose B-FIRE, a binning-free diffusion implicit neural representation framework for hyper-accelerated MR reconstruction capable of reflecting instantaneous 3D abdominal anatomy. B-FIRE employs a CNN-INR encoder-decoder backbone optimized using diffusion with a comprehensive loss that enforces image-domain fidelity and frequency-aware constraints. Motion binned image pairs were used as training references, while inference was performed on binning-free undersampled data. Experiments were conducted on a T1-weighted StarVIBE liver MRI cohort, with accelerations ranging from 8 spokes per frame (RV8) to RV1. B-FIRE was compared against direct NuFFT, GRASP-CS, and an unrolled CNN method. Reconstruction fidelity, motion trajectory consistency, and inference latency were evaluated.

Problem

• 4DMRI
• motion-resolved MRI
• undersampled k-space
• instantaneous dynamics
• motion artifacts

Innovation

• binning-free
• diffusion model
• implicit neural representation
• hyper-accelerated MRI
• motion-resolved reconstruction