Research On The Key Methods Of Hepatocellular Carcinoma Assisted Diagnosis System Based On CT

According to statistics of investigations and researches, liver cancer is threatening the human health seriously, it is not only the main type of cancers, but also a liver disease with highest lethality. Among all the types of liver cancers, the hepatocellular carcinoma is the most serious. The reason why it has the worst treatment effect is the delays of definite diagnose and treatment. Misdiagnoses always exist although there are multiple diagnostic methods of hepatocellular carcinoma. Thus, computer assisted diagnostic of medical images emerges; it could be deemed as a part of digital image process and mode recognition, and it can provide the quantification diagnose in the general investigation of disease and provide a referenced assistive diagnostic basis; such technology has a profound future in the imaging diagnosis of diseases with positive significance.This paper studies deeply the domestic and foreign scholars on the auxiliary liver disease diagnosis theory and research method, summarized and absorbed the model of image processing in the medical industry application, at the same time for HCC CT images of computer-aided diagnosis method has carried on the study and research, through the computer image processing technology for liver CT image feature extraction, feature selection and classification comparison, for HCC auxiliary diagnosis method based on CT were improved. The main work has the following four points:1. In view of the current liver texture feature vector extraction method was studied, not only limited to extract only in the past to study and put forward "Great influence", such as commonly used based on gray level co-occurrence matrix of the four feature vector, but based on the extraction of texture feature vector to carry on the comprehensive, including six major characteristics, the gray histogram based on gray level co-occurrence matrix is not commonly used eight characteristics and based on the five characteristics of gray trip matrix of three aspects, a total of 27 eigenvectors are extracted for texture analysis.2. On the way of feature selection, the reference more often using principal component analysis, in view of the principal component analysis to extract the main component of ambiguity, a different director ingredients extracted from different defects, this paper proposes a improved principal component analysis (PCA), on the extraction of the characteristics of the original features of linear combination into the original filter, select the practical significance and meaning of the original feature vector for the training of the classifier.3. On the classification using support vector machine (SVM) method to construct classifier, adopt the method of All samples, random sampling and the method of cross validation for training and validation, respectively, and using the same data source to contrast with experimental research to improve the former method.4. The constructed visual user interface, demonstrated the HCC in the form of auxiliary diagnosis technology used in the future. Interface including CT image preprocessing, image area to intercept suspicious, effective characteristics evaluated preliminary diagnosis and CT images, and other functions. In the future, after the large database building interface will provide more convenient service for medical workers.Regarding to the test result, the accuracy rate of judging hepatocellular carcinoma and normal liver respectively are 92.9%. Above 89.2% accuracy before improvement. Simulation experiment result shows that the system of hepatocellular carcinoma diagnosis and visualized interface described in the thesis has a certain degree of realizability. The system could provide the referable quantification information to medical workers and has a positive significance in the research of medical images assistive diagnosis.