Combination of Iterative Metal Artifact Reduction and Virtual Monoenergetic Reconstruction Using Split-Filter Dual-Energy CT in Patients With Dental Artifact on Head and Neck CT

Abstract

BACKGROUND. Head and neck CT can be limited by dental hardware artifact. Both postprocessing-based iterative metal artifact reduction (IMAR) and virtual monoenergetic imaging (VMI) reconstruction in dual-energy CT (DECT) can reduce metal artifact. Their combination is poorly described for single-source DECT systems. OBJECTIVE. The purpose of this study was to compare metal artifact reduction between VMI, IMAR, and their combination (VMIIMAR) in split-filter single-source DECT of patients with severe dental hardware artifact. METHODS. This retrospective study included 44 patients (nine woman, 35 men; mean age, 66.0 ± 10.4 years) who underwent head and neck CT and had severe dental hardware artifact. Standard, VMI, IMAR, and VMIIMAR images were generated; VMI and VMIIMAR were performed at 40, 70, 100, 120, 150, and 190 keV. ROIs were placed to measure corrected attenuation in pronounced hyperattenuating and hypoattenuating artifacts and artifact-impaired soft tissue and to measure corrected artifact-impaired soft-tissue noise. Two radiologists independently assessed soft-tissue interpretability (1-5 scale), and pooled ratings were analyzed. Readers selected the preferred reconstruction for each patient. RESULTS. Mean hyperattenuating artifact-corrected attenuation was 521.0 HU for standard imaging, 496.4-892.2 HU for VMI, 48.2 HU for IMAR, and 32.8-91.0 HU for VMIIMAR. Mean hypoattenuating artifact-corrected attenuation was -455.1 HU for standard imaging, -408.5 to -679.9 HU for VMI, -37.3 for IMAR, and -17.8 to -36.9 HU for VMIIMAR. Mean artifact-impaired soft tissue-corrected attenuation was 10.8 HU for standard imaging, -0.6 to 24.9 HU for VMI, 4.3 HU for IMAR, and -2.0 to 7.8 HU for VMIIMAR. Mean artifact-impaired soft tissue-corrected noise was 58.7 HU for standard imaging, 38.2 to 129.7 HU for VMI, 11.0 HU for IMAR, and 5.8 to 45.6 HU for VMIIMAR. Median soft-tissue interpretability was 1.2 for standard imaging, 1.1-1.2 for VMI, 3.7 for IMAR, and 2.0-3.8 for VMIIMAR. Artifact-impaired soft tissue-corrected attenuation and soft-tissue interpretability significantly improved (p < .05) for VMIIMAR versus IMAR only at 100 keV. The two readers preferred VMIIMAR at 100 keV in 56.8% and 59.1% of examinations. CONCLUSION. For reducing severe artifact due to dental material, IMAR has greater effect than VMI. Though the results for IMAR and VMIIMAR were similar overall, VMIIMAR had a small benefit at 100 keV. CLINICAL IMPACT. VMI and IMAR techniques in split-filter DECT may be combined for clinical head and neck imaging to reduce artifact from dental hardware and improve image quality.