Spatial frequency-dependent signal-to-noise ratio as a generalized measure of image quality

Abstract

ABSTRACT A generalized, objective image quality measure can be defined for X-ray based medical projection imaging: the spatial frequency-dependent signal-to-noise ratio SNR = SNR(u,v). This function includes the three main image quality parameters, i.e. spatial resolution, object contrast, and noise. The quantity is intimately related to the DQE concept, however its focus is not to characterize the detector, but rather the detectability of a certain object embedded into a defined background. So also effects from focus size and radiation scatter can be quantified by this method. The SNR(u,v) is independent of basic linear post-processing steps such as appropriate windowing or spatial filtering. The consideration of the human visual system is beyond the scope of this concept. By means of this quantity, different X-ray systems and setups can be compared with each other and with theoretical calculations. Moreover, X-ray systems (i.e. detector, beam quality, geometry, anti-scatter grid, basic linear post-processing steps etc.) can be optimized to deliver the best object detectability for a given patient dose. In this paper SNR(u,v) is defined using analytical formulas. Furthermore, we demonstrate how it can be applied with a test phantom to a typical flat panel detector system by a combination of analytical calculations and Monte Carlo simulations. Finally the way this function can be used to optimize an X-ray imaging device is demonstrated. Keywords: Medical X-ray imaging, image quality measure, DQE , contrast, noise, spatial resolution 1. INTRODUCTION For the optimization of the setup of a medical projection X-ray imaging device it is always a challenge to find a good compromise between image sharpness and noise level