Research On Particle Swarm Optimization In Medical Micrograph Segmentation

Image segmentation is an important step in target detection and identification is the basic work of many advanced image processing technology. Also, Medical image segmentation is the foundation and difficulty in image processing and analysis. In the computer-assisted semen analysis system, the segmentation and recognition of the sperm target, DNA fragmentation and sperm morphology is the key follow-up step. So far, a variety of image segmentation methods have been proposed. Thresholding techniques have become an important method of image segmentation because it is simple and easy to achieve. However, most of threshold selection methods need exhaustive search, especially when asked to find the best threshold combination, to achieve multi-threshold segmentation, they become time-consuming, inefficient and unable to meet the application requirements.In order to quickly and accurately find the best combination for multi-threshold, we should find an efficient algorithm to search for the best threshold combination, which can meet the accuracy, real-time requirements.Particle Swarm Optimization is a new global search strategy based on population optimization, is a typical intelligent optimization algorithum.PSO has been proved to be an effective global optimization algorithm. It is easy to implement, quickly convergence, and has been successfully applied to many engineering fields. However, the standard Particle Swarm Optimization is easy to fall into local optimum, and slow convergence. In this paper, the advanced PSO is applied to medical micrograph segmentation, and the analysis results are compared. Specifically as follows:(1) Assisted Sperm Analysis System is introduced, also the background and significance of this study. And the existing image segmentation methods are summarized, the main work in this paper is described.(2) The principle and the research progress of PSO are introduced. And prior to some issues of this method such as easily trapped in local, slow convergence, the standard PSO is improved. A new method is proposed that not only adaptive adjusts inertial weight according to the evolutionary rate, but also through mutation when the particles trapped in stagnant state so that to search other regions of space. Meanwhile, an experiment of a large number of different test functions is did, experimental results show the improved method can speed up the converge speed, and effectively defend from local extremum.(3) The improved PSO is combined with Otsu, the maximum entropy method and fuzzy C-mean clustering methods, and used in medical image segmentation, including single-threshold, and multi-threshold segmentation, the results are compared. The results show that, this method can quickly and accurately find the best combination of the segmentation threshold, providing a solid theoretical foundation and experimental data for practical application.