Methodology for Estimation of Tissue Noise Power Spectra in Iteratively Reconstructed MDCT Data

Abstract

Iterative reconstruction algorithms have been recently introduced into X-ray computed tomography imaging. Enabling patient dose reduction by up to 70% without affecting image quality they deserve attention; therefore properties of noise present in iteratively reconstructed data should be examined and compared to the images reconstructed by conventionally used filtered back projection. Instead of evaluating noise in imaged phantoms or small homogeneous regions of interest in real patient data, a methodology for assessing the noise in full extent of real patient data and in diverse tissues is presented in this paper. The methodology is based on segmentation of basic tissues, subtraction of images reconstructed by different algorithms and computation of standard deviation and radial onedimensional noise power spectra. Tissue segmentation naturally introduces errors into estimation of noise power spectra; therefore, magnitude of segmentation error is examined and is considered to be acceptable for estimation of noise power spectra in soft tissue and bones. As a result of this study it can be concluded that iDose4 hybrid iterative reconstruction algorithm effectively reduces noise in multidetector X-ray computed tomography (MDCT) data. The MDCT noise has naturally different characteristics in diverse tissues; thus it is object dependent and phantom studies are therefore unable to reflect its whole complexity.