Artificial Intelligence–Driven Ultra-Fast Superresolution MRI

Abstract

Magnetic resonance imaging (MRI) is the keystone of modern musculoskeletal imaging; however, long pulse sequence acquisition times may restrict patient tolerability and access. Advances in MRI scanners, coil technology, and innovative pulse sequence acceleration methods enable 4-fold turbo spin echo pulse sequence acceleration in clinical practice; however, at this speed, conventional image reconstruction approaches the signal-to-noise limits of temporal, spatial, and contrast resolution. Novel deep learning image reconstruction methods can minimize signal-to-noise interdependencies to better advantage than conventional image reconstruction, leading to unparalleled gains in image speed and quality when combined with parallel imaging and simultaneous multislice acquisition. The enormous potential of deep learning-based image reconstruction promises to facilitate the 10-fold acceleration of the turbo spin echo pulse sequence, equating to a total acquisition time of 2-3 minutes for entire MRI examinations of joints without sacrificing spatial resolution or image quality. Current investigations aim for a better understanding of stability and failure modes of image reconstruction networks, validation of network reconstruction performance with external data sets, determination of diagnostic performances with independent reference standards, establishing generalizability to other centers, scanners, field strengths, coils, and anatomy, and building publicly available benchmark data sets to compare methods and foster innovation and collaboration between the clinical and image processing community. In this article, we review basic concepts of deep learning-based acquisition and image reconstruction techniques for accelerating and improving the quality of musculoskeletal MRI, commercially available and developing deep learning-based MRI solutions, superresolution, denoising, generative adversarial networks, and combined strategies for deep learning-driven ultra-fast superresolution musculoskeletal MRI. This article aims to equip radiologists and imaging scientists with the necessary practical knowledge and enthusiasm to meet this exciting new era of musculoskeletal MRI.

Updates on Compositional MRI Mapping of the Cartilage: Emerging Techniques and Applications

“…Reviewed papers were selected from Google Scholar and PubMed, searching for the words: "quantitative," "MRI," and "cartilage," between 2020 and 2023. To reduce overlap with recent review articles on rapid knee MRI 48 and DL methods for fast imaging and relaxometry, 49,50 some references already in these reviews may not be cited here. This section provides an update on the newly emerging techniques described after the ones cited in these previously published review articles.…”

A Review of Advanced Multifunctional Magnetic Nanostructures for Cancer Diagnosis and Therapy Integrated into an Artificial Intelligence Approach

“…MHNs have proven to be effective as both (T1 and T2 MRI contrast) in terms of both their size and surface functionalization as long as both are optimally tuned. Moreover, artificial intelligence (AI) is well suited to MRI due to its inherent soft-tissue contrast, variety of structural and physiological acquisition protocols, and diagnostic capabilities [ 135 ]. Notably, MRI will transform into a new era of quantitative imaging with AI by utilizing these large data structures to revolutionize its largely qualitative clinical applications [ 136 ].…”

Physics‐informed deep learning for T2‐deblurred superresolution turbo spin echo MRI

“…Deep learning superresolution (SR) is a promising approach to reduce MRI scan time without requiring custom sequences or iterative reconstruction. [1][2][3][4][5][6][7][8][9] Previous deep learning SR approaches have been primarily applied to through-slice SR (e.g., "DeepResolve" 1,9 ) or to gradient-echo pulse sequences, [3][4][5]7 for which lowering the nominal resolution at the scanner can be reasonably modeled by relaxation-independent resolution degradation models such as slice profile changes or k-space truncation. However, lowering the in-plane spatial resolution of other common pulse sequences such as turbo spin echo (TSE) 10 also involves modifying TEs and echo trains.…”