The influence of the learning set on the evaluation of microcalcifications using artificial neural networks

Abstract

Neural networks have been devised to detect the location of microcalcifications (segmentation method) to evaluate the degree of malignancy, based on identified features describing potential microcalcifications. Previous studies focused on the selection of the optimal vector of features (set of features), describing the mammogram and evaluation of the MLBP and Fahlman’s neural networks as decision support tools. This article presents next study of applications of the neural network methods for the automatic analysis of mammograms. Methods elaborated for the digitalization of the mammograms are presented earlier. The study of identification and classification of masses and microcalcifications in digital mammograms are shown. The purpose of this paper is to indicate the representativeness of the dataset on the results obtained using artificial neural networks in the detection of microcalcifications.