Scientific Session 15 — Musculoskeletal - CTWednesday, May 3, 2017
2883. Effects of Reconstruction Planes on CT Image Quality When Using Iterative Reconstruction Techniques
Ott J, Verdun F, Omoumi P, Rotzinger D, Becce F*. Institute of Radiation Physics, University of Lausanne, Lausanne, Switzerland
Address correspondence to F. Becce (firstname.lastname@example.org)
Objective: Although the effects of iterative reconstruction (IR) techniques on CT image quality and noise reduction have extensively been evaluated in the axial plane, their effects on the objective image quality of thin coronal and sagittal reformations have not been investigated to our knowledge. Furthermore, most previous studies on IR techniques did not take into account their nonlinearity, which introduces a dependency of the image contrast and noise on spatial resolution. Therefore, the objective of our study was to evaluate the effects of image reconstruction planes on objective CT image quality when using IR techniques.
Materials and Methods: Conventional and custom CT image quality phantoms were scanned on a 64-MDCT scanner. Acquisition parameters were the same as for various musculoskeletal CT examinations, including unenhanced knee CT and CT arthrography: 120 kVp; 144 mAs; volume CT dose index, 11.5 mGy. Data were reconstructed in the axial, coronal, and sagittal planes using filtered back projection (FBP), 40% and 80% adaptive statistical IR (ASIR), and model-based IR (MBIR) techniques by applying the following parameters: FOV, 20 × 20 cm; slice thickness, 0.625 mm; interval, 0.625 mm, and sharp (bone) convolution kernel except for MBIR (detail). Modulation transfer function (MTF) and noise power spectra were calculated. Target transfer functions (TTF), an object-specific physical metric, were also obtained for three different contrast materials reproducing CT attenuation numbers (at 120 kVp) of cortical bone (polytetrafluoroethylene), fat (polyethylene), and articular cartilage (poly[ethyl methacrylate]). Finally, we computed the detectability indexes (d') of 1-mm-diameter articular cartilage lesions using an updated nonprewhitening with eye (NPWE) model observer.
Results: In the axial plane, noise power spectra significantly decreased when switching from FBP to ASIR and MBIR. On coronal and sagittal reformations, there was a strong noise reduction compared with axial images. However, MBIR still helped further reduce noise. No significant changes in spatial resolution (MTF) were found when using IR techniques. However, spatial resolution decreased when switching from axial to coronal and sagittal reconstruction planes. On axial images, TTF revealed increased spatial resolution for the two lower-contrast materials (poly[ethyl methacrylate] and polyethylene) when using ASIR and particularly MBIR. However, TTF remained unchanged and maximal for the high-contrast material (polytetrafluoroethylene) independently of image reconstruction techniques. On coronal and sagittal reformations, polytetrafluoroethylene exhibited higher spatial resolution than poly[ethyl methacrylate] and polyethylene. We also noted that ASIR did not alter spatial resolution, whereas MBIR significantly increased spatial resolution for all three contrast materials. Finally, NPWE detectability indexes decreased on coronal and particularly sagittal reformations compared with axial images. There was a drastic increase in detectability in all three reconstruction planes when using MBIR.
Conclusion: Task-based physical metrics are required to accurately evaluate CT image quality when using IR techniques. CT images reconstructed in coronal and sagittal planes have a significantly lower image quality with potential effects on diagnostic performance. IR techniques, particularly MBIR, help partially compensate for the loss of diagnostic information.