Research Of Detection Algorithm Of Pulmonary Nodules Based On DICOM Sequence

Cancer is one of the primary disease to threaten to human life. Because of the property of a high incidence,high mortality and high incidence of new cases,lung cancer become the most deadly cancer disease. With the change of people’s habits and environmental degradation,the population of lung cancer is increasing,thus social concerns is also increasing. Due to increasing the level of development of medical and medical inputs, and updating CT, MRI and other equipment constantly, the number of the piece CT images also increase. The way of early detection and early treatment can improve survival of lung cancer, but doctors’ workload and high detection rate is contradictory. Computer-aided diagnosis system can greatly reduce the duplication to assisting doctors to increase speed and efficiency of detection.The sign of lung cancer symptoms on image is pulmonary nodules. Material used in this paper is LIDC dataset, namely the United States the authority of the lung cancer research data sets. According to pulmonary nodules of medical imaging characteristics and properties in this paper,use image processing and data mining technology in medical processing and studying about image sequence. According to histogram of lung CT images, the paper uses mean threshold, denoising, contour extraction and morphological operations to complete lung parenchyma segmentation, which denoising method is based on profile and area. Then, it proposes and uses hybrid space enhancement algorithm and based on gaussian model and hessian matrix multi-scale circular enhanced filtering algorithm for lung CT images to rise gray level of suspected pulmonary nodules, including hybrid space nodule enhancement algorithm enhances lung nodule contour information to achieve regional effects, and multi-scale circular enhancement filter enhances suspected pulmonary nodule based on round modeling and inhibit vascular tissue to achieve enhancement of pulmonary nodules area, then highlight the performance results. Edge detection based on the Live-wire algorithm and region growing algorithm is proposed and used to achieve rapid extraction of parenchyma of sequence images. Then with the statistical characteristics of the nature of pulmonary nodules and improved genetic algorithm is used to optimizing the combination of features to improve lung nodule detection rate. Finally, based on two class features group of RBF kernel SVM classification algorithm is used to determine pulmonary nodules. The accuracy of pulmonary nodule detection is that detecting pulmonary nodules possess the proportion of all lung nodules, while the false positive rate is the average number of false-positive nodules in each sequence. Through a series of image processing and data analysis process, the ultimate goal of this paper is to improve the rate and accuracy of detection of pulmonary nodules and reduce the false-positive rate.At present, at home and abroad, lung CAD system to a large extent is in the course of the study, and not is applied in the actual hospital. In this paper, through testing, verifying functional performance of systems, and comparing with domestic and foreign research software systems, there are the advantage of high-speed, high stability, high sensitivity and high specificity.