Automatic segmentation of masses in digital mammograms using particle swarm optimization and graph clustering

This paper presents a methodology for automatic segmentation of masses in digital mammograms based on two principles: thresholding and evolutionary algorithm. As the staring point of the particles of the swarm, we used Otsu. Then, we applied the Particle Swarm Optimization (PSO) to optimize, evolutionarily, the search for the global maximum of the thresholds in order to achieve a better segmentation. After the segmentation stage, we executed a reduction of false positives based on region growing, area filter and Graph Clustering.