RESUMO
PURPOSE: To evaluate magnetic resonance imaging (MRI) artifacts near metallic spinal instrumentation using both conventional metal artifact reduction sequences (MARS) and 3D multispectral imaging sequences (3D-MSI). MATERIALS AND METHODS: Both MARS and 3D-MSI images were acquired in 10 subjects with titanium spinal hardware on a 1.5T GE 450W scanner. Clinical computed tomography (CT) images were used to measure the volume of the implant using seed-based region growing. Using 30-40 landmarks, the MARS and 3D-MSI images were coregistered to the CT images. Three independent users manually segmented the artifact volume from both MR sequences. For five L-spine subjects, one user independently segmented the nerve root in both MARS and 3D-MSI images. RESULTS: For all 10 subjects, the measured artifact volume for the 3D-MSI images closely matched that of the CT implant volume (absolute error: 4.3 ± 2.0 cm3 ). The MARS artifact volume was â¼8-fold higher than that of the 3D-MSI images (30.7 ± 20.2, P = 0.002). The average nerve root volume for the MARS images was 24 ± 7.3% lower than the 3D-MSI images (P = 0.06). CONCLUSION: Compared to 3D-MSI images, the higher-resolution MARS images may help study features farther away from the implant surface. However, the MARS images retained substantial artifacts in the slice-dimension that result in a larger artifact volume. These artifacts have the potential to obscure physiologically relevant features, and can be mitigated with 3D-MSI sequences. Hence, MR study protocols may benefit with the inclusion both MARS and 3D-MSI sequences to accurately study pathology near the spine. LEVEL OF EVIDENCE: 2 J. Magn. Reson. Imaging 2017;45:51-58.