Pulmonary Nodule Detection And Recognition Algorithm Based On Ct Images

In recent years, the lung cancer becomes one of the severest malignant tumors dangerous for human's health, and it is the cancer in lowest survival rate after it is diagnosed. If the lung cancer can be detected in the early time, the quality of life and survival rate of patients can be improved. Early lung cancer is a form of pulmonary nodules in the body, the lung nodules are similar to pulmonary vascular cross-section that is similar to round. The doctors even the clinical experience ones may miss diagnosing the disease. As the large volume of data with CT images, one patient often has hundreds of images, it is easy to make a wrong diagnosis. Computer-aided diagnosis system is based on medical knowledge, combined with modern computer technology, image processing and artificial intelligence theory. It can finish the mark of the suspected area automatically. It increases doctors'efficiency and reduces the rate of misdiagnosis and missed diagnosis rate.According to the related research of the lung nodule detection algorithm, this paper propose a new lung nodule detection algorithm, the specific work of the article is as follows:1 In this paper, the overall situation adaptive threshold is used for the image segmentation of lung parenchyma. Complete removing of the part of torso by seeking the best threshold repeatedly and extract the reign of lung parenchyma by boundary tracing.2 Aimed at the situation that the time of fuzzy C-means clustering algorithm is too long and the segmentation effect of image is not ideal, the improved algorithm is posed. The algorithm is improved based on combining the statistical properties of histogram and optimizing the membership. On this basis, complete the research of the FCM algorithm's initial parameter to get the parameter option to fit lung parenchyma image processing. Extract the Region of Interest by improved fuzzy C-means algorithm, the result proves that the algorithm has good real-time and good segmentation results.3 Choose the characteristics of pulmonary nodules, and then propose the Support Vector Machine (SVM) based on single Eigen value optimization and optimize eigenvectors. Make a corresponding study to the option of normalized way and Kernel function on the basis of the principle of increasing the accuracy. The simulation result shows that the algorithm has higher accuracy.