Research On MRI Image Segmentation Algorithm Based On Pulse Coupled Neural Network

During the process of diagnosis,doctors often need the assistance of medical images,accurate medical image segmentation is of great significance for doctors to diagnose diseases and make treatment plans for some diseases.Magnetic resonance imaging(MRI)is one of the most widely used brain imaging methods,which can obtain medical images from the brain by using magnetic resonance phenomenon.However,the special transmission mode of MRI will cause the image to be interfered by mixed noise,and because of the low contrast of the medical image itself and the shape of the internal tissue is uncertain,it will be more difficult to separate the focus from the normal tissue.In order to design an automatic MRI brain image segmentation algorithm to meet the practical requirements of clinical medicine,in this paper,the MRI image in the image database of Harvard Medical School are used to study and process from two aspects of denoising and segmentation,which achieve good results.The work of this paper is summarized as follows:Firstly,in order to remove the noise interference,the improved wavelet transform algorithm is used to denoise the MRI brain image.In the low frequency part,thresholds are set according to the distribution of wavelet coefficients for signal-to-noise separation,and the wavelet coefficients in the non-target region are weakened by filtering.In the high frequency part,the coefficients are processed by the adaptive threshold function.The experimental results show that the improved denoising algorithm achieves better visual effects and denoising effect evaluation parameters compared with the common denoising algorithms in medical images.Then,in order to ensure the segmentation effect and real-time performance,this paper adopts the PCNN model to segment the MRI brain images,and uses the improved GSO algorithm to optimize the parameters.The concept of adaptive step size was introduced in the improved GSO algorithm so that the step size of individual fireflies could be adjusted automatically with the number of iterations.At the same time,the algorithm marks the optimal population value and classifies fireflies in each iteration process.According to the type of fireflies,the individual movement mode is set to three kinds.Finally,in order to successfully apply the improved GSO algorithm to MRI brain images,an appropriate fitness function was set to regulate the movement of fireflies during the firefly iteration process.In this paper,the characteristics of MRI brain images are analyzed,and the commonly used evaluation parameters are listed.In order to eliminate the inaccuracy of image segmentation caused by a single parameter,the weighted sum of the two parameters is determined as the final fitness function through comparative experiments.After the improvement of firefly algorithm and fitness function,the segmentation experiment of denoised MRI images was carried out.The experimental results show that the improved GSO algorithm is superior to other swarm intelligence algorithms in terms of optimization and convergence,and it is not easy to fall into the local optimal solution through the optimization experiment on the special test function of swarm intelligence algorithm.Compared with other commonly used segmentation methods in MRI images,the improved algorithm proposed in this paper can effectively reduce the over-segmentation phenomenon while retaining the inner edge features of the image,and achieve better segmentation results.