Pulmonary Nodule Segmentation Algorithm Based On Finite Mixture Model

Lung cancer is one of the most important malignant tumors that seriously endanger human health.It is the first cancer in China with incidence and mortality.Early diagnosis of lung cancer can increase the five-year survival rate of lung cancer patients.CT imaging is an important method for the early diagnosis of lung cancer.However,with the increase of the CT image data volume,the imaging physician's reading workload increases,resulting in an increase in missed diagnosis rate and misdiagnosis rate of lung cancer.The computer aided detection and diagnosis(CAD)system used in CT can reduce the imaging doctor's reading time,provide necessary auxiliary diagnosis information,and improve the diagnosis efficiency and accuracy.Pulmonary nodules are important manifestations of early lung cancer in CT images.In the CAD system,the accurate segmentation of pulmonary nodules is an important prerequisite for reducing the rate of missed lung cancer detection and realizing the judgment of benign and malignant pulmonary nodules.This topic discusses the existing pulmonary nodule segmentation algorithm,analyzes its existing deficiencies,and does the following research on lung nodule segmentation methods for chest CT images: First,the segmentation method of lung parenchyma is studied.Segmentation of lung parenchyma is an important step in the segmentation of lung nodules,which can effectively reduce the search range of lung lesions and improve detection efficiency.After preliminary processing,image binarization,removal of surrounding tissue,and separation of left and right lung parenchyma,the primary segmentation of the lung parenchyma is achieved.The contour of the lung parenchyma is repaired by the roller ball method,and the segmentation of the lung parenchyma is completed.Then,the pulmonary nodule segmentation method based on the single function finite mixture model is studied,including Gaussian mixture model and Gamma mixture model.This topic gives a new model parameter estimation method,namely adaptive particle swarm optimization(APSO),which realizes the simultaneous optimization of the number of model components and parameters in the segmentation process,and obtains a more accurate pulmonary nodule segmentation result.Finally,based on the fact that the gray distribution curve of normal lung parenchyma and pulmonary nodules does not necessarily obey a single distribution functions,pulmonary nodule segmentation is achieved using a combination of multiple distribution functions.This topic gives a specific set of distribution functions,and presents a lung nodule segmentation method based on a self-selected mixed distribution model.The selection of the optimal distribution function model is completed,the parameters of the model are estimated and the quantity of the mixture is optimized,then the segmentation of pulmonary nodules is realized.In addition,in view of the superiority of the model,a pulmonary nodule segmentation method based on the self-selected mixed distribution model of neighborhood information is studied.It uses pixel neighborhood information to reduce the influence of noise on the segmentation result and further improves the accuracy of the lung nodule segmentation.The lung nodule segmentation method described in this paper is validated and analyzed experimentally,and the evaluation method of lung nodule segmentation based on the finite mixture model is given.Comparing this algorithm with other similar lung nodule segmentation methods,the experimental results show that the segmentation method of pulmonary nodules described in this paper has a better segmentation effect.