Total Knee Arthroplasty MRI Featuring Slice-Encoding for Metal Artifact Correction: Reduction of Artifacts for STIR and Proton Density–Weighted Sequences

Abstract

The purpose of this article is to compare slice-encoding for metal artifact correction (SEMAC) sequences versus optimized standard MRI sequences in patients with total knee arthroplasty (TKA). Forty-two patients with TKA underwent 1.5-T MRI. Sequences optimized for metal implant imaging (SEMAC) were compared with standard sequences optimized with high bandwidth for STIR and proton density (PD)-weighted images. In 29 patients, CT was available as reference standard. Signal void and insufficient fat saturation were quantified. Qualitative criteria (anatomy, distortion, blurring, and noise) were assessed on a 5-point scale (1, no artifacts; 5, severe artifacts) by two readers. Abnormal imaging findings were noted. A Student t test and a Wilcoxon signed rank test was used for statistics. Signal void areas and insufficient fat saturation were smaller for the SEMAC sequences than for the optimized standard sequences (p ≤ 0.005 for all comparisons). Depiction of anatomic structures was better on STIR with SEMAC versus standard sequences optimized with high bandwidth (score range, 2.9-3.7 vs 4.2-4.9) and on PD-weighted imaging with SEMAC versus standard sequences optimized with high bandwidth (score range, 2.5-3.5 vs 3.1-3.8), which was statistically significant (p < 0.001 to p = 0.007 for different structures). Distortion and noise were lower for SEMAC than for the standard sequences (p ≤ 0.001), whereas no technique had a clear advantage for blurring. Detection of abnormal imaging findings was markedly increased for the SEMAC technique (p < 0.001) and was most pronounced for STIR images (98 and 74 findings for STIR with SEMAC for readers 1 and 2, respectively, vs 37 and 37 findings for readers 1 and 2, respectively, for STIR with standard sequences optimized with high bandwidth). Sensitivity for detection of periprosthetic osteolysis was improved for STIR with SEMAC (100% and 86% for readers 1 and 2, respectively) compared with STIR with standard sequences optimized with high bandwidth (14% and 29% for readers 1 and 2, respectively). SEMAC sequences showed a statistically significant artifact reduction. The detection of clinically relevant findings such as periprosthetic osteolysis was markedly improved.