Research Of The Prediction On The Model To Distinguish The Benign And Malignant Solitary Pulmonary Nodules Based On The Hybrid Imaging

Recently, Lung diseases have been one of the most major factors that affectthe quality even the human life, due to the exacerbating problems of diet,environmental problem and other problems. Therefore, the early detection andtreatment of the lung disease has become the most effective way of preventionand the treatment of lung cancer, also a hot area of the current research. In theearly stage of the disease, the imaging manifestations of the lung disease aresolitary pulmonary nodules (SPNs).What’s more, there is a great differencebetween the signs of the malignant and benign solitary pulmonary nodules. Sothe distinguishing of the malignant and benign of the solitary pulmonary nodulesbased on the different characteristics has been an important way to identify thecharacter of the solitary pulmonary nodules.The medical diagnosis of lung cancer based on the hybrid imaging ofPET/CT technology has not only considered the clinical characters of SPNs butalso integrated the signs of PET and CT images, which overcome theshortcomings of diagnosis information based on the single image. Currently, theearly identification of benign and malignant solitary pulmonary nodules is stilldependent on the physician’s reading piece experience and the characters weusually used cannot be quantified, there will inevitably be missed, andmisdiagnosed cases. In order to reduce the misdiagnose and missed diagnosisphenomenon as much as possible, we need to quantize all the signs of the solitary pulmonary nodules and analyze the relationships between the signs andthe characteristic of solitary pulmonary nodules, and finally construct a model toprediction the malignant and benign of the solitary pulmonary nodules.In order to construct the prediction model of the benign and malignant ofthe solitary pulmonary nodules, this paper mainly includes the followingaspects:1. An improved dual-fuzzy support vector machine distinguishing benignand malignant nodules. The ultimate objective of the diagnose is todistinguish the malignant and benign of SPNs, the traditionalclassification method of SPNs based on SVM considered that all thesamples have the same contributions to getting the optimal hyperlane,and without taking into account of the impact of the relationshipbetween the samples. In this paper, an improved fuzzy support vectormachine to classify the SPNs based on the different characters wasproposed, in which we combined the signs on the PET images and CTimages and classified the SPNs based on the clinical characters.2. A mathematical model to predict the benign and malignant SPNs. Themethod first quantified all the characters of the samples, and thenanalyzed the relationship between every single character and thecharacteristics of SPNs based on the single factor analysis. Finally weuse the multivariate logistic analysis to analyze the clinical data of SPNsin CT and PET images and based on which constructing a regressionequation to predict the malignant and benign of the SPNs.At the end of the dissertation, experiments were carried out to evaluate theperformance of the methods described in this dissertation. The experimentalresults strongly supported the effectiveness of the method, which improved theaccuracy of the discrimination of the benign and malignant solitary pulmonarynodules, while reducing the rate of misdiagnosis.