Comprehensive assessment of nonuniform image quality: Application to imaging near metal
In this work we present an ensemble of methods for quantifying spatial variations in MR image quality and demonstrate its utility for benchmarking imaging performance near metallic implants.
MR image quality depends on a balance of many factors including scan time, signal-to-noise ratio, spatial resolution, tissue contrast, and artifacts. The remarkable flexibility of this modality permits a myriad of possibilities for compromise across these quality factors in the service of a given clinical application. On the other hand, this flexibility underscores the immense technical challenge for benchmarking imaging performance.
The application of MR imaging near metal presents an exemplary case study for the advancement of image quality quantitation methods. The magnetic susceptibility mismatch between soft tissue and metal creates a large, static magnetic field perturbation causing artifacts with a high degree of spatial variation. State-of-the-art multispectral imaging techniques (e.g., SEMAC, MAVRIC-SL) can mitigate some of these artifacts, but also can induce further quality variations in terms of SNR and spatial resolution.
In this work, we propose a new paradigm for the comprehensive analysis of image quality near metallic implants. A modular phantom design is demonstrated to support repeatable toggling of both local structure and susceptibility artifact to support the co-registered volumetric quantitation of artifact, spatial resolution, and SNR with complete ground truth information. Spatial resolution mapping is enabled by a novel method based on local point spread function estimation and shown to benefit from the use of a gyroid lattice for the phantom structure. Results are demonstrated with 2D-FSE and MAVRIC-SL pulse sequences. The phantom is designed to be reproducible at low-cost using entry-level rapid prototyping equipment and nontoxic materials.
Toolbox for Non-uniform Image Quality Analysis (TONIQ): All materials needed to reproduce or adapt this work are publicly available here: https://github.com/artoews/toniq
Toews AR, Lee PK, Nayak KS, Hargreaves BA. Comprehensive assessment of nonuniform image quality: Application to imaging near metal. Magn Reson Med. 2024;1-15. doi: 10.1002/mrm.30222
Pictorial summary of the image quality analysis paradigm. The modular phantom is imaged in four different configurations (panel A) to quantify 3D spatial variations in image quality across four measures (panels B-E): intensity artifact, SNR, geometric distortion, and spatial resolution. Each displayed image/map represents a data volume in three spatial dimensions. Five repetitions of a single target protocol are used to acquire all inputs except the Structured-Plastic input in (E) which uses a higher resolution protocol as reference.
