SU-G-206-10: Low-Contrast Detectability Vs. Dose for CT Images Reconstructed Using Filtered Backprojection and Iterative Reconstruction: Assessment with a Model Observer

• Published in: Medical physics (Lancaster) 2016-06, Vol.43 (6), p.3641-3641
• Main Authors: Favazza, C, Ferrero, A, McMillan, K, Bruesewitz, M, Yu, L, Leng, S, Kofler, J, McCollough, C
• Format: Article
• Language: English
• Subjects: Biomedical modeling; Computed tomography; Image detection systems; Image reconstruction; Image scanners; Medical image contrast; Medical image reconstruction; Three dimensional image processing; Three dimensional sensing
• ISSN: 0094-2405; 2473-4209
• DOI: 10.1118/1.4956951

Purpose: To evaluate CT low-contrast detectability at different dose levels in images reconstructed using filtered backprojection (FBP) and iterative reconstruction (IR). Methods: A validated channelized Hotelling observer (CHO) model was used to detect low contrast objects in images of a helical CT phantom (Model 061, CIRS) using an Aquilion Prime 160 scanner (Toshiba Medical Systems). Twenty-one different objects of 7 diameters (2.4, 3.2, 4.6, 6.3, 9.5, and 10 mm) and 3 contrast levels (∼10, 15, and 25 HU) were evaluated. Images were acquired using 5 different tube current settings (45, 60, 90, 140, and 180 mA) and reconstructed using both FBP and IR (Adaptive Iterative Dose Reduction (AIDR) 3D). For each tube current, the phantom was scanned 100 times, yielding sets of 100 object-present and object-absent images at the same x-y locations. All images were evaluated using the CHO and detectability indices (d’) for each object were calculated at all dose levels. Results: For the same tube current setting, images formed using IR yielded d’ values on average 8% higher than images reconstructed using FBP. When dose was reduced by more than 25% of the full dose, however, results showed reduced low-contrast detectability in AIDR-3D images compared to 100% dose, FBP images. With the vendordefault settings, dose is automatically reduced by 70-75% when AIDR 3D reconstructions are selected, which our data indicate would significantly reduce low contrast detectability. Conclusion: At higher dose levels (minimal dose reduction), AIDR-3D improved object detectability as compared to full-dose FBP images. However, when dose was reduced by more than 25%, object detectability was substantially reduced for images reconstructed with AIDR-3D compared to full-dose FBP images. CH McCollough is supported by industry funding from Siemens Healthcare