Research And Implementation Of Liver Cancer Identification Based On Liver Statistical Appearance Model And Optimized SVM

With the rapidly development of the digital medical equipment, image processing, pattern recognition, Computer aided diagnosis (Multimate Principal Component Analysis, MPCA) technology accured at the historic moment of the study of meidical images. The rate of liver cancer increased year by year in China, There are more demands on computer aided diagnosis of liver cancer, so the development of the liver cancer image analysis and recognition technology has important research significance and application value.To protect the special space structure information of liver images, in this paper we research two models based on Multiple linear principal component analysis (Multimate Principal Component Analysis, MPCA) method, and generalized N dimensional principal component analysis (Generalized N-Dimensional Principal Component Analysis, GND-PCA) method according to the reference on the basis of a large number of literature in China and abroad. then used ant colony optimization method for classification. First we extracted the Texture feature, fractal feature to establish the high-order tensor of liver volume, then used two method to establish the statistical appearance model, And then used the method of dimension reduction mehtod to map the model into the ant colony optimized SVM (Ant Colony Optimize-Support Vector Vector, ACO-SVM, ACO-SVM) to identification liver cancer.In order to verify the effectiveness of the algorithm, Liver cancer identification experiment was carried out based on the liver images, the experimental results show that:The two tensor model is better than the original vector model, and GND-PCA method is more superior to MPCA method in both classification and running rate, this algorithm can achieve better classification results in the classification of liver disease and the multiple classification problems, it can also accurately assist the doctor in the diagnosis of liver cancer.