Research And Implementation Of Pancreatic Cancer Detection Based On CT Images

According to the World Health Organization, it reported that pancreatic cancer is a highly malignant tumor which ranked sixth in cancer deaths. It is less than 5% survival rate of in five years. In recent years, the incidence rate increased year by year. Pancreatic cancer is difficult to diagnose early, once confirmed, the average survival time is normally only six months. The clinical features are to develop fast, short course and high mortality. Therefore, early detection, early diagnosis and early treatment for pancreatic cancer mortality have decreased significance and clinical value. As high resolution and low injury for body, CT (computed tomography) image is considered to be a highly sensitive and valuable diagnostic method which has been the early detection of pancreatic cancer. It can also reflect the position of pancreatic pathology accurately and vividly. So it favors by the doctor. But a large number of CT images will add significant burden on doctors. With digital image processing, pattern recognition technology continuing to evolve, it is the biomedical problems to be solved for developing a set of objective, reliable and non-invasive method for early diagnosis of pancreatic cancer CT images.On the basis of research results which refer extensively at home and abroad, this paper takes the establishment of quantum genetic algorithm to optimize the support vector machine (QGA-SVM) classification model and proposes a method based on CT images of pancreatic cancer detection. This test method is divided into training and testing stages of learning practice stages, including the region of interest (ROI) selection, texture feature extraction, feature selection, classifier design. First, it manually segmented region of interest (the pancreas area) from the abdominal CT images; Then, extract texture feature; Then for the removal of redundant information, feature selection is selected in classes spacing; Finally it is divided by support vector machine classification. To improve the performance of the classifier, it starts from the nuclear parameters, which is using quantum genetic algorithm parameters by optimized SVM model and establish QGA-SVM classification model.In this paper, one hundred and fifteen cases of training data set and test data sets were randomly selected. Through multiple randomized trials, it shows that this algorithm is feasible for the detection of pancreatic cancer. Compared with the BP neural network, the traditional SVM, genetic algorithm optimization SVM (GA-SVM), it proves that the establishment of QGA-SVM classification model takes obvious advantages of achieving higher classification accuracy and provides valuable reference for clinical diagnosis for doctors.