Computer-aided Method Of Detecting Lung Fissure Based On CT Images

The paper studies the way to realize the segment and test of pulmonary fissure from lung CT images with the support of computer. It also conducts an in-depth study of relevant methods to test pulmonary fissure in medical images. Through all-round analysis, the study finds that the supervised method could effectively segment and test the pulmonary fissure. Currently, the wide varying range of pixel values, border ambiguity and distortion, along with surrounding vessels, image noises and other factors, have jointly created difficulties in distinguishing each region of the lung. In order to test the pulmonary fissure, the study, on the basis of unsupervised test theory, utilizes supervised methods to segment and test the pulmonary fissure. In the unsupervised pulmonary fissure test, the Hessian Matrix and the Gaussian Filter are utilized to enhance the pulmonary fissure of the lung image without supervision. The theory proposing no-supervision serves as the foundation of supervised test. And in the supervised pulmonary fissure test, the Gaussian Filter, Hessian Matrix and Gradient technique are adopted in the first step to obtain 57 feature images. To get more valid feature images of pulmonary fissure, a feature selection procedure will be conducted among these feature images. The mainly utilized method for features selection among the 57 feature images is the Sequential Floating Forward Selection, thus getting more valid feature images concerning pulmonary fissure. Then the study utilizes the Linear Discriminant Classifier, Quadratic Discriminant Classifier, Support Vector Machine, SVM (containing Gaussian kernel) as well as the k-Nearest Neighbor Classifier in the experiment. The experiment with the support of several classifiers finds the proper pulmonary fissure type of KNN classifier along with the best value of classification. In this way, the supervised test of pulmonary fissure could be obtained. In addition, the paper also proposes a two-phase supervised test to get even clearer images of pulmonary fissure test. Scale spacing, dynamic programming and other methods are used to segment and separate pulmonary fissure in the end.Methods proposed by the paper can be adopted in the simulation experiments of multiple groups of clinical images to test pulmonary fissure, which could help doctors to analysis disease and segment the lung lobe. The test and segment of pulmonary fissure plays a significant role in the study of medical images.