Naveen Subhas MD, Presenter: Research Grant, Siemens AG

Sahar Shiraj MD, Abstract Co-Author: Nothing to Disclose

Andrew Primak PhD, Abstract Co-Author: Employee, Siemens AG

Joshua Matthew Polster MD, Abstract Co-Author: Nothing to Disclose

Andreas Krauss PhD, Abstract Co-Author: Employee, Siemens AG

Jean Pierre Schils MD, Abstract Co-Author: Nothing to Disclose

Joseph Iannotti, Abstract Co-Author: Consultant, Johnson & Johnson Consultant, Zimmer Holdings, Inc Consultant, Tornier, Inc Royalties, Johnson & Johnson Royalties, Zimmer Holdings, Inc Royalties, Tornier, Inc Royalties, Biomet, Inc Royalties, MTF Royalties, Wolters Kluwer nv Royalties, Reed Elsevier

Iterative metal artifact reduction (IMAR) is a new sinogram inpainting technique to reduce CT metal artifact which adds high frequency data to improve visualization close to metal edges. Our purpose was to compare the image quality and accuracy of attenuation values near hardware of IMAR and standard filtered back projection (FBP) in patients with shoulder arthroplasties (SA).

8 patients (6 male, avg age 60) with 9 SAs were scanned on a FLASH CT (Siemens) with a standard protocol (140 kVp, 300 eff mAs, 0.6mm collimation, eff pitch 0.35-0.8). Images were reconstructed on a standalone workstation with a smooth kernel (B30) and 0.6mm slice thickness. 3 IMAR reconstructions with different amounts of high frequency data: IMAR (least), IMAR1.5 (more), IMAR2.5 (most) and FBP were ranked for image quality by 5 readers in a side by side comparison from best=1 to worst=4 for bone, soft tissue, metal-bone interface and overall quality. Accuracy of attenuation near hardware was quantified as the absolute difference (AD) between avg HU within a region of interest (ROI) near hardware and for an ROI containing similar tissues on a slice without hardware.

IMAR1.5 was ranked best for humeral cortex (avg 1.4), glenoid trabeculae (avg 1.36) and glenoid cortex (avg 1.4). IMAR2.5 was ranked best for humeral trabeculae (avg 1.2). IMAR was ranked the best for deltoid muscle (avg 1.2). IMAR1.5 and 2.5 were ranked best for metal-bone interface (avg 1.3). FBP was ranked worst for all structures (avg 3.38 -3.49). All readers ranked IMAR1.5 and 2.5 over FBP (p<0.05) for overall image quality. 3 readers preferred IMAR1.5 and 2.5 over IMAR (p<0.05). There were no significant differences between IMAR1.5 and 2.5 for any reader. Accuracy of attenuation near hardware was significantly better in all 3 IMAR reconstructions than FBP for bone (AD 86–99 vs 430, p<0.006) and soft tissue (AD 23–29 vs 450, p=0.003). There was no significant difference between the IMAR techniques in bone (p=0.4) or soft tissue (p=0.4).

IMAR, especially with added high frequency data, had superior image quality and more accurate attenuation values near hardware than standard FBP in patients with shoulder arthroplasties.

IMAR is a promising new CT technique to reduce metal artifact that is fully automatic and computationally inexpensive and has the potential to replace standard FBP in patients with hardware.

Subhas, N, Shiraj, S, Primak, A, Polster, J, Krauss, A, Schils, J, Iannotti, J, Novel CT Metal Artifact Reduction Prototype for Evaluation of Shoulder Arthroplasties.  Radiological Society of North America 2013 Scientific Assembly and Annual Meeting, December 1 - December 6, 2013 ,Chicago IL. http://archive.rsna.org/2013/13022210.html