Research On Medical Image Classification Based On Twin Support Vector Machine

Breast X-ray photography is a common method for early diagnosis of breast cancer, the method mainly depends on the doctor’s observation and analysis of medical images. But medical images contain massive information, it is difficult to find the hidden useful information by naked eye. With the development of Internet technology, online diagnosis becomes inevitable, doctors will face massive data. So we use data mining method to analyze the information that is hard to find by naked eye to help doctors make a diagnosis, which has become the research hotspot recently. So efficient data mining methods are introduced for rapid and accurate diagnosis and identification of medical images, so as to avoid doctor’s wrong judgment and improve the doctor’s work efficiency.Support vector machine is a machine learning method that established on statistical learning theory. It has a unique advantage in solving the non-linear and small sample size problem. Twin support vector machine solves two smaller quadratic programming problems to get a pair of nonparallel hyperplanes, so that samples in each class are close to the corresponding hyperplane and be away from the other. It is a fast classification method that has the advantages of traditional support vector machine and strong data processing and generalization ability, which is widely used in pattern recognition and data classification. In this paper, aiming at the problem of fuzzy medical image in multi class classification, we deeply discuss and study two kinds of medical image classification methods based on twin support vector machine, which are used for multi class classification to improve the classification accuracy of medical image. The main work is as follows:(1) Decision tree twin support vector machine based on genetic algorithm for multi class classification algorithm, and its application in classification of medical image.In this paper we propose a decision tree twin support vector machine algorithm based on genetic algorithm for multi class classification. The algorithm combines the decision tree and twin support vector machine to construct the classifier to solve the multi class classification problem, which overcomes the fuzzy problem of traditional twin support vector machine for multi class classification. The method builds decision tree with feature data by genetic algorithm to separate the fuzzy region of samples to improve the sample recognition rate. For each node of the decision tree we use twin support vector machine to train classifier, and finally we use the trained classifier for classification and prediction. The new algorithm has been applied in multiple sets of UCI data set and medical image classification, the experimental results show that the new algorithm has higher classification accuracy and rapid training speed than traditional twin support vector machine.(2) Decision tree twin support vector machine based on kernel principal component analysis for multi class classification algorithm, and its application in medical image classification.For a complete description of the information contained in medical image, we usually need to extract massive features. So the high dimension of the input space may make the computational complexity grows exponentially with the increasing sample dimension, so we must use the effective method to reduce the dimension of input space. We propose a decision tree twin support vector machine based on kernel principal component analysis algorithm for the realization of data dimension reduction, and then we use the proposed algorithm to train classifier with the feature dataset after dimension reduction to improve the performance of the classifier. The new algorithm is applied in medical image classification, the experimental results show that the new algorithm has better classification results than decision tree twin support vector machine.