Abstract— Computer-aided diagnosis (CADx) of mammographic microcalcifications (MCs) is designed to reduce the number of false biopsies by increasing the positive predictive value of the radiologist’s interpretation. Although several algorithms were developed in the last two decades, the performance of these systems remains unsatisfactory and the need for developing efficient automated feature extraction and selection techniques is still high. In our attempt to address this demand, we propose a morphology-based CADx system for which we extract a set of 44 morphological features that describe the shape and the distributions of microcalcifications. In this paper, we present a heuristic model selection algorithm using a PSO-SVM framework that combines feature selection and SVM performance optimization steps. We also compare the performance of the feature selection using a binary PSO method against the outweighed nested-subsets method. To validate the proposed feature extraction and model selection methods, two datasets of microcalcification (MC) clusters have been used: the mini-MIAS and a digital mammography dataset from Bronson Methodist Hospital in Kalamazoo, Michigan. The obtained results demonstrate the effectiveness of the proposed CADx and indicate that a PSO-SVM framework using a binary PSO feature search method is more powerful than using an outweighed nested-subsets method.