Quantitative Modelling of Microcalcification Detection in Digital Mammography

Abstract

Abstract. This article presents a simulation framework for the image acquisition on digital mammography systems. The framework is used to analyse the performance of a previously developed method for the detection of microcalcifications by a series of top-hat operators. The framework allows to determine the theoretical number of false positives and true positives for a given set of acquisition parameters and size of microcalcifications. A minimal size of microcalcification that can be detected with sufficient certitude is analysed as a function of the pixel size of the detector and the acquisition conditions. An improved selection scheme for the detection threshold is developed which uses the model and the acquisition parameters to obtain a near-optimal detection for a wide range of acquisition conditions. Finally the model is evaluated on clinical images.