Research On Medical Image Fusion Technology Based On Pulse Coupled Neural Network

The development of computer technology has continuously promoted the advancement of medical technology.Medical imaging is an indispensable auxiliary tool for guiding doctors' diagnosis and treatment.Because of the different imaging principles,single-modal medical images can only focus on displaying unilateral information.For example,CT images can display bone information more clearly,and MRI images are more detailed for soft tissue expression.Therefore,doctors often need to pass two or more images for diagnosis.Image fusion technology not only combines the key information carried by medical images of different modalities,but also provides better visual effects.At present,there are many researches on the fusion methods of medical images at home and abroad,such as multi-scale decomposition and neural network fusion methods.Among them,the neural network has better data analysis and computational ability to provide better fusion effect.The pulse-coupled neural network is derived from the cat's visual cortex because of its design principle.Therefore,the network can simulate the mammalian optic nerve cortex mechanism for image processing.Therefore,when applied to image fusion,good results have been achieved.But precisely because of its unique network structure,the use of parameters and the determination of the number of iterations have become a problem that hi nders its continued progress.In view of the existing problems of medical image fusion algorit hms based on pulse coupled neural network,this paper has carried out the following research:(1)By analyzing the defects of traditional PCNN network,by estimating the comparison ratio,an approximate passive period of passive PCNN neurons is obtained.The error betwee n the estimated passive period and the actual passive period is analyzed and proved.The initia l phase of the start pulse of the neuron is derived,and a stable initial phase passive pulse perio d is given.The setting of the parameters and the setting of the appropriate number of iterations have served as a reference for comparison.(2)Based on a simplified S-PCNN model,the parameters are set appropriately and the number of iterations is standardized.A new fusion method is proposed.Firstly,the source ima ge is decomposed into high and low frequency coefficients by non-subsampled shear wave tra nsform.Then,the high frequency coefficients are merged by using the intersecting visual corti cal model and the improved simplified pulse coupled neural network model..The low freque ncy is fused by the regional gradient energy method.Finally,the fused high and low frequenc y coefficients are obtained by the inverse non-subsampled shear wave transform to obtain the fused image.The fusion method exhibits a fusion effect superior to the contrast method both i n subjective and objective aspects.(3)Based on the fusion method proposed in this paper and the improved simplified puls e coupled neural network,the medical image fusion system is realized.The system includes f unctions such as image input,method selection,image fusion and fusion image output.It is better to facilitate the fusion of medical images and optimize the fusion representation effect.