Low contrast detectability and spatial resolution with model-based Iterative reconstructions of MDCT images: a phantom and cadaveric study.

OBJECTIVES: To compare image quality [low contrast (LC) detectability, noise, contrast-to-noise (CNR) and spatial resolution (SR)] of MDCT images reconstructed with an iterative reconstruction (IR) algorithm and a filtered back projection (FBP) algorithm. METHODS: The experimental study was performed on a 256-slice MDCT. LC detectability, noise, CNR and SR were measured on a Catphan phantom scanned with decreasing doses (48.8 down to 0.7 mGy) and parameters typical of a chest CT examination. Images were reconstructed with FBP and a model-based IR algorithm. Additionally, human chest cadavers were scanned and reconstructed using the same technical parameters. Images were analyzed to illustrate the phantom results. RESULTS: LC detectability and noise were statistically significantly different between the techniques, supporting model-based IR algorithm (p < 0.0001). At low doses, the noise in FBP images only enabled SR measurements of high contrast objects. The superior CNR of model-based IR algorithm enabled lower dose measurements, which showed that SR was dose and contrast dependent. Cadaver images reconstructed with model-based IR illustrated that visibility and delineation of anatomical structure edges could be deteriorated at low doses. CONCLUSION: Model-based IR improved LC detectability and enabled dose reduction. At low dose, SR became dose and contrast dependent. KEY POINTS: • Model- based Iterative Reconstruction improves detectability of low contrast object. • With model- based Iterative Reconstruction, spatial resolution is dose and contrast dependent. • Model-based Iterative Reconstruction algorithms enable improved IQ combined with dose-reduction possibilities. • Improvement of SR and LC detectability on the same IMR data set would reduce reconstructions