Objective assessment of low-contrast computed tomography images with iterative reconstruction.

Abstract Objective This study aims to assess low-contrast image quality using a low-contrast object specific contrast-to-noise ratio (CNR LO ) analysis for iterative reconstruction (IR) computed tomography (CT) images. Methods A phantom composed of low-contrast rods placed in a uniform material was used in this study. Images were reconstructed using filtered back projection (FBP) and IR (Adaptive Iterative Dose Reduction 3D). Scans were performed at six dose levels: 1.0, 1.8, 3.1, 4.6, 7.1 and 13.3 mGy. Objective image quality was assessed by comparing CNR LO with CNR using a human observer test. Results Compared with FBP, IR yielded increased CNR at the same dose levels. The results of CNR LO and observer tests showed similarities or only marginal differences between FBP and IR at the same dose levels. The coefficient of determination for CNR LO was significantly better (R 2  = 0.86) than that of CNR (R 2  = 0.47). Conclusion For IR, CNR LO could potentially serve as an objective index reflective of a human observer assessment. The results of CNR LO test indicated that the IR algorithm was not superior to FBP in terms of low-contrast detectability at the same radiation doses.