Smart Pathological Brain Detection System By Weighted K Nearest Neighbor Based On Hybrid Genetic Algorithm And Particle Swarm Optimization Algorithm

With the continuous development of modern advanced science and technology,and the continuous progress of medical technology and medical methods,combining traditional medicine with modern science had become one of the important tasks of the development of the times.This thesis confirmed the background and significance of the study.After that,we made a brief analysis and introduction to the research and development of brain magnetic resonance image at home and abroad.This thesis put forward one kind of smart pathological brain detection,the main research was medical image processing technology of computer aided diagnosis.The aim was to make a good classification of medical images and to distinguish between the brain images of the sick and the healthy.In this thesis,the magnetic resonance imaging of the brain was classified by using the idea of image feature extraction and classification optimization,and obtained the following research results:1.The method of data augmentation is used to rotate the image of healthy brain to increase the number of healthy brain images and solve the problem of unbalanced data.In order to extract the feature more better in the future research.The images of brain MRI after data augmentation was extracted by wavelet entropy.I reduced the number of features from 65536 to 7,as the most important part of the input data of the W-kNN classifier in the future experiment.2.The Weighted k-Nearest Neighbor was used as the classification algorithm in this thesis.It is a variant of the kNN algorithm.W-kNN distributes weight according to the importance of sample characteristics.The appropriate weight of automatic learning,assigning different importance to different features.The introduction of weight can improve the classification performance of kNN algorithm,and the classification accuracy and stability are better than that of the kNN algorithm.3.This thesis proposed a hybrid algorithm based on genetic algorithm and particle swarm optimization algorithm,combining the two algorithms by learning from each other's strong points.The idea of cross and mutation of GA was introduced into the standard PSO algorithm to avoid the premature convergence of the algorithm and lead to the local optimal solution.The seven optimal solutions obtained by the algorithm were used as the weight of the sample features,and the input data of the weighted kNN algorithm were obtained by multiplying the sample features.Then 3-fold cross validation was used to train the data to ensure the training effect of the data set.Finally,the quality of the experiment was judged by the classification accuracy.It can be found that the hybrid optimization weighted K nearest neighbor algorithm based on genetic algorithm and particle swarm optimization(WE-W-kNN-GA PSO)had the best classification effect.The dataset were tested,and the classification accuracy reached 96.97%.In the future,I will make greater efforts for the research of SPBD.