Modeling of multiple technical factors effect on noise in computed tomography

Abstract

The aim of this work was to obtain a predictive statistical model for pixel noise on water phantom CT images, with appropriate physical meaning of model parameters. A statistical model was obtained using images from the head and abdominal sections of a physical phantom, which includes as independent variables: peak kilo voltage, tube current per scanning time, phantom diameter, collimation, reconstruction diameter, reconstruction mode, reconstruction filters and post processing filters, interaction between scanned object diameter and reconstruction diameter was also taken into consideration. The adjusted coefficient of multiple determination for noise model was 0.98 (p-value < 0.001). The model also described correctly the influence of each experimental factor, from which weighting functions were developed. A reference noise can be computed from model in order to test appropriateness of system state for noise predictions. The present predictive image-noise model provides quantitative estimation of effect on noise according to several setting strategies. © 2009 Springer-Verlag.