Solitary Pulmonary Nodule Diagnosis Model Research Based On PET-CT

In recent years, the incidence of lung cancer and other lung diseases is rising. Meanwhile, the mortality of lung cancer has also been much higher than other cancers. Most of the lung cancer at early stage is manifested as a solitary pulmonary nodule. It’s the early stages that the a large number of lung tomography images and doctors professional level in diagnostic process, the limitations of observed visually is the leading cause of the high rate of misdiagnosis and missed diagnosis. In the current solution to the diagnosis of lung canc er, the main diagnostic method of treating lung cancer is the use of computer radiographic image data analysis, modeling, and improve diagnostic efficiency and accuracy physician. PET/CT tomography technique is most effective tool in the early detection of solitary pulmonary nodule. Between accurate segmentation of lung parenchyma and classification of solitary pulmonary nodules is the most effective step to improve the cure rate of lung cancer. In this paper, the both current research status are summarized, and its improved method is proposed. Under the background of PET-CT imaging technology development in the clinical diagnosis of lung cancer, the main research work from the following several aspects in the design of the diagnosis model of solitary pulmonary nodules:1. For the complex lung and the chest area containing pleural nodules in clinical diagnosis, region growing method cannot be accurately included pleural nodules and better preserve the edge information characteristics of lung co ntour in the process of lung parenchyma segmentation. On the basis of preliminary lung parenchyma segmentation by region growing method, this paper proposes a semiautomatic lung parenchyma repair method based on the rolling ball method. This method can effectively repair the defects of the lung parenchyma, improve the segmentation accuracy of lung parenchyma, and the lung parenchyma after the patch is in agreement with the similarity of the lung parenchyma of the physician manual segmentation. It can preserve the edge information of the lung profile, which provides an important basis for the later solitary pulmonary nodule detection, and greatly reduces the rate of missed diagnosis of lung cancer.2. This paper establish a high correlation and low redundancy optimization feature sub set on the basis of the function and structure of solitary pulmonary nodule image visual feature, fully combined with the clinical manifestations of solitary pulmonary nodules, by analyzing the correlation between the various features and nodules of benign or malignant. For support vector machine(SVM) exist problems in the solitary pulmonary nodules benign and malignant classification, such as fault points and samples near the optimal classification surface, the same contribution of each sample and without considering the classification surface affected by the correlation between samples. This paper proposed Solitary pulmonary nodule diagnosis model design based on the SVM- KNN in PET-CT to achieve noninvasive diagnosis, Combining Support Vector Machines and K-nearest neighbor classification algorithm, based on the current rules of classification of benign and malignant nodules diagnosis, using multidimension features information and taking into consideration of the correlation of the equivocal sample and the near sample point. By validating the classification performance of the diagnostic model. Under the premise for the improvement of the classification accuracy, the method of this paper reduces the false positive rate of the detection and has a high sensitivity. It provides efficient, objective and convenient auxiliary means for solitary pulmonary nodules diagnosis and has a positive effect on the early detection of lung cancer.