Data and Physics driven Deep Learning Models for Fast MRI Reconstruction: Fundamentals and Methodologies

TL;DR

This paper reviews recent data and physics-driven deep learning techniques for accelerating MRI scans, highlighting methodologies, challenges, and future directions to improve image quality and reconstruction speed.

Contribution

It provides a comprehensive overview of recent advances in MRI acceleration methods combining data-driven and physics-based models, including new algorithms and integration strategies.

Findings

Advances in algorithm unrolling and generative models for MRI reconstruction

Integration of hardware acceleration techniques like parallel imaging

Discussion of challenges such as data heterogeneity and model generalization

Abstract

Magnetic Resonance Imaging (MRI) is a pivotal clinical diagnostic tool, yet its extended scanning times often compromise patient comfort and image quality, especially in volumetric, temporal and quantitative scans. This review elucidates recent advances in MRI acceleration via data and physics-driven models, leveraging techniques from algorithm unrolling models, enhancement-based methods, and plug-and-play models to the emerging full spectrum of generative model-based methods. We also explore the synergistic integration of data models with physics-based insights, encompassing the advancements in multi-coil hardware accelerations like parallel imaging and simultaneous multi-slice imaging, and the optimization of sampling patterns. We then focus on domain-specific challenges and opportunities, including image redundancy exploitation, image integrity, evaluation metrics, data heterogeneity, and model generalization. This work also discusses potential solutions and future research directions, with an emphasis on the role of data harmonization and federated learning for further improving the general applicability and performance of these methods in MRI reconstruction.