Improved Visualization of Juxtaprosthetic Tissue Using Metal Artifact Reduction Magnetic Resonance Imaging: Experimental and Clinical Optimization of Compressed Sensing SEMAC.

OBJECTIVES

The purpose of this study was to identify an optimal imaging protocol for metal artifact reduced magnetic resonance imaging by application of different imaging and postprocessing parameters in compressed sensing slice-encoding for metal artifact correction (CS-SEMAC) and to test it in patients with total hip arthroplasty (THA).

MATERIALS AND METHODS

In an experimental setup, a phantom consisting of a standard THA embedded in gadolinium-containing agarose was scanned at 1.5 T. Pulse sequences included coronal short tau inversion recovery (STIR), T1-weighted (w), and T2-w CS-SEMAC sequences. All pulse sequences were acquired with 11, 19, and 27 slice-encoding steps (SESs), respectively. For each raw dataset, postprocessing was performed with variations of the parameters: (1) number of iterations (5, 10, 20, 30, 50) and (2) normalization factor (0.0005, 0.001, 0.002, 0.003, 0.005). Following, in clinical magnetic resonance scans of patients with THA, identical STIR, T1-w, and T2-w pulse sequences with 11 and 19 SESs were acquired and were postprocessed similarly with variations in parameters. Semiquantitative outcome measures were assessed on a 5-point scale (1 = best, 5 = worst). The overall best image quality was determined. Signal-to-noise ratio and contrast-to-noise ratio were calculated. Statistical analyses included descriptive statistics, t-tests, multivariate regression models, and partial Spearman correlations.

RESULTS

Scan times varied between 2:24 (T2-w, 11 SESs) and 8:49 minutes (STIR, 27 SESs). Reconstruction times varied between 3:14 minutes (T1-w, 11 SESs, 5 iterations) and 85:00 minutes (T2-w, 27 SESs, 50 iterations). Signal-to-noise ratio and contrast-to-noise ratio increased with increasing SESs, iterations, and normalization factor. In phantom scans, artifact reduction was optimal with an intermediate normalization factor (0.001) and improved with higher SESs and iterations. However, iterations greater than 20 did not improve artifact reduction or image quality further. On the contrary, ripple artifacts increased with higher SESs and iterations. In clinical scans, up to 20 iterations reduced blurring of the image; no further reduction was observed with iterations greater than 20. A normalization factor of 0.001 or 0.002 was best for reduction of blurring, whereas the soft tissue contrast was better and the distortion of soft tissue was less severe with lower normalization factors. Overall best soft tissue image quality was found for STIR and T1-w images with 19 SESs, 10 iterations, and a normalization factor of 0.001, and for T2-w images with 11 SESs, 10 iterations, and a normalization factor of 0.0005.

CONCLUSIONS

Optimized advanced acceleration and reconstruction algorithms of CS-SEMAC have been identified to reduce metal artifacts in patients with THA enabling imaging with clinically feasible acquisition and reconstruction times.