Solitary Pulmonary Nodules Detection Based On Hybird Imaging

In the field of modern medicine, cancer is always a significant threat to human life and health. Among various cancers, the incidence and the mortality of lung cancer are on top. In China, lung cancer is often difficult to find, or when patients suffering from lung cancer is usually already in the middle and late. The early detection of cancer lesions is the key to its cure. Usually early lung cancer presents as solitary pulmonary nodule, solitary pulmonary nodule due to its smaller radius easily overlooked by physician. On the other hand, benign and malignant solitary pulmonary nodules judge usually requires pathological examination, the physician simply by imaging is difficult to determine its benign or malignant. The computer-aided diagnosis system, able to obtain more diagnostic information from the image than the physician, provided the conditions for the judgment of benign and malignant solitary pulmonary nodules imaging. This paper presents a hybrid imaging based on solitary pulmonary nodule segmentation and classification method that can effectively detect patients with pulmonary solitary pulmonary nodules, provided the conditions for subsequent computer-aided benign and malignant judgment.This article relies on "Hybrid imaging computer-aided diagnosis of solitary pulmonary nodules program,"the National Natural Science Foundation of China. In this paper, based on the idea experienced physician diagnosis,use of hybrid imaging information through interactive segmentation and detection of nodules reached. Of this study is divided into two areas look. The first is suspicious nodule segmentation work, will probably be the solitary pulmonary nodule region separated from the image. Followed by suspicious nodules classification, segmentation of suspicious nodules extract features for classification, to determine whether it is nodules.Suspicious nodule segmentation stage, you first need to preprocess the CT images, the paper chose the enhanced diffusion filter denoising images, remove noise generated and stored when the image generated. The method can effectively retain edge information and internal parenchymal nodules. Post to remove noise in CT images of the lung parenchyma segmentation region, we use dynamic segmentation method based on optimal threshold, the method has a higher coding efficiency. After obtaining the lung parenchyma, the coordinate obtained by mapping the relationship between a PET image and a CT image registration method for interworking based on the mapping to the lung region in the CT image in a PET PET in the lung parenchyma. Then using the PET images presented in this paper based on variable -sized template matching method, get suspicious nodule region, the area is transformed into the seed point sequence. Coordinate mapping information through registration obtained seed points will be mapped into CT images. Finally, the improved algorithm feng shui Ridge CT image segmentation in precise areas suspicious nodules After comparing the results with manual segmentation physician’s verification, to the suspicious nodule segmentation method proposed by the segmentation results with manual results are highly consistent.After getting suspicious nodule region, the paper texture feature extraction and metabolic characteristics of suspicious nodules composed of feature vectors area of machine learning classification to determine whether a suspicious nodule solitary pulmonary nodules. Which uses a CT feature extraction method for calculating GLCM, PET SUV extract the mean and maximum. Selected based on SVM classifier RBF kernel function in machine learning. Experimental results show that the proposed method can effectively reduce the guaranteed sensitivity on the basis of false positives.