A multi-objectively-optimized graph-based segmentation method for breast ultrasound image

Segmentation of medical image, as the most essential and important step in the computer-aided diagnosis system, can greatly influence the system performance. Better segmentation to a great extent means better performance. Among many proposed segmentation algorithms, graph-based segmentation has become a hot one in the past few years because of the simple structure and rich theories. After the robust graph-based segmentation method (RGB) was introduced in 2010, a parameter-automatically-optimized robust graph-based segmentation method (PAORGB) was presented in 2013 as well, to optimize the two key parameters of RGB utilizing the particle swarm optimization algorithm (PSO). However, single-objectively-optimized PAORGB cannot well guarantee the global optimization. Therefore, this paper continues the work of PAORGB and proposes a multi-objectively-optimized robust graph-based segmentation method (MOORGB) to further improve the performance of RGB. Experimental results have shown that MOORGB can get better segmentation results from breast ultrasound images compared to PAORGB.