Objective: The purpose of this study was to validate iterative reconstruction technique in oncologic chest computed tomography (CT).
Methods: An anthropomorphic thorax phantom with 4 simulated tumors was scanned on a 64-slice CT scanner with 2 different iterative reconstruction techniques: one model based (MBIR) and one hybrid (ASiR). Dose levels of 14.9, 11.1, 6.7, and 0.6 mGy were used, and all images were reconstructed with filtered back projection (FBP) and both iterative reconstruction algorithms. Hounsfield units (HU) and absolute noise were measured in the tumors, lung, heart, diaphragm, and muscle. Contrast-to-noise ratios (CNRs) and signal-to-noise ratios (SNRs) were calculated.
Results: Model-based iterative reconstruction (MBIR) increased CNRs of the tumors (21.1-192.2) and SNRs in the lung (-49.0-165.6) and heart (3.1-8.5) at all dose levels compared with FBP (CNR, 1.1-23.0; SNR, -7.5-31.6 and 0.2-1.1) and with adaptive statistical iterative reconstruction (CNR, 1.2-33.2; SNR, -7.3-37.7 and 0.2-1.5). At the lowest dose level (0.6 mGy), MBIR reduced the cupping artifact (HU range: 17.0 HU compared with 31.4-32.2). An HU shift in the negative direction was seen with MBIR.
Conclusions: Quantitative image quality parameters in oncologic chest CT are improved with MBIR compared with FBP and simpler iterative reconstruction algorithms. Artifacts at low doses are reduced. A shift in HU values was shown; thus, absolute HU values should be used with care.