Research On Maximum Entropy Multi-Threshold MRI Image Segmentation Based On Improved Particle Swarm Optimization

Nuclear magnetic resonance imaging is one of the most widely used and mature medical imaging methods.The segmentation technology of NMR image has great application value in medical diagnosis,lesion localization,tissue determination,biological experiments and other fields.Threshold segmentation algorithm is the main method of nuclear magnetic resonance image segmentation at present.Based on the traditional threshold image segmentation algorithm,the improved particle swarm optimization algorithm is used to optimize multi-threshold parameters.A new segmentation algorithm for nuclear magnetic resonance image is proposed.Compared with the traditional single threshold segmentation algorithm,it has the characteristics of high efficiency,fast speed and high segmentation accuracy.The specific contents of the article include:(1)Taking the maximum entropy threshold segmentation algorithm and the maximum inter-class threshold segmentation algorithm as examples,the threshold image segmentation algorithm is discussed.Based on the traditional threshold segmentation algorithm,an improved particle swarm optimization based maximum entropy threshold segmentation algorithm for NMR images is proposed.(2)By optimizing the inertia weight,shrinkage factor and other parameters,an expansion model is designed and introduced to improve the particle swarm optimization algorithm.The convergence performance of the improved particle swarm optimization algorithm,the basic particle swarm optimization algorithm and the quantum particle swarm optimization algorithm are compared through the test function,and the convergence speed and search performance of the new algorithm are verified.(3)A sample of 30 cases(150 images)of human brain MRI sequences from the Nuclear Magnetic Resonance Database of the Center of Developing Brain Semnar in Imperial College London was taken.the maximum entropy multi-threshold segmentation algorithm based on the improved particle swarm optimization is tested and the performance of the algorithm is verified.(4)Nuclear magnetic resonance image segmentation technology has very high potential application value in animal research field,but related research has published less literature and there is no open image database.In order to verify the segmentation effect of the new algorithm for small animal brain magnetic resonance images,the pigeon brain magnetic resonance images obtained in the preliminary work of the laboratory were taken as the object,and the pigeon brain,cerebellum,medulla oblongata,optic tectum and other tissues were taken as the segmentation targets.The image segmentation experiments of the new algorithm were carried out with the reference of the standard library established by artificial segmentation.A preprocessing algorithm for pigeon brain magnetic resonance image is designed by combining linear gray enhancement,median filtering and area comparison.