PCNN Medical Image Segmentation Based on PSO

Medical image segmentation is a vital step in medical processing, but subject to its imaging factors, medical image has a low contrast, similar grayscale value and fuzzy boundary and it is impossible for naked eyes to accurately recognize the boundaries of each region. Under this circumstance, to segment the medical image accurately and quickly has become a research hotspot and difficulty. With a background of biology, pulse coupled neural network (PCNN) is proposed based on the synchronization pulse firing phenomenon of the cerebral cortex of such animals as cats and monkeys and referred to as the third-generation artificial neural network, it has many features such as dynamic neuron, space-time characteristics, automatic spread of wave and synchronization pulse firing. PCNN is a mono-layer model neural network and it needs no training for achieve pattern recognition, image segmentation and target classification and so on. Meanwhile, PCNN transfers twodimensional space variables into one-dimensional time pulse sequences while processing the image. In this way, PCNN model is one step closer to the actual biological neural network and it surely has stronger ability to process the input information and better performance. The basic idea of particle swarm optimization (PSO) is to search the optimal solution through the collaboration and information sharing among the individuals and it has the characteristics of evolutionary computation and swarm intelligence. Due to the firing acquisition feature of PCNN neurons, the image segmentation based on PCNN is an image segmentation method based on the strength and similarity of the image pixels and it has the characteristics of adaptive image segmentation. However, its binary segmentation on the image greatly weakens the hierarchy of image and it is bad for the subsequent image processing. Therefore, this paper improves the pulse generator of PCNN with PSO and proposes PCNN multi-value segmentation method based on PSO. The experiment result shows that the method of this paper can effectively adjust PCNN parameters and preserve the geometrical characteristics of the image. It not only maintains PCNN's excellent characteristics on medical image segmentation, but it also effectively retains the hierarchy of the medical image itself.