The Discrimination System Of Solitary Pulmonary Nodules On CT Images

Lung Cancer has been one of the most malignant tumors which have the most damagingto human life，and mortality rate is higher than any other cancer and increasing year by year．The mortality of lung disease is increasing year by year because of the deterioration of theenvironment and a large increasing in smoking population.Early detection of cancer is themost effective method for the treatment.In the diagnosis of lung diseases,Medical imagingapplication is very extensive Which includes the X line imaging,CT and MRI examinationand so on.CT is considered the best imaging examination of lung disease.This paper is mainlydevoted to the Research on computer aided diagnosis of the lung nodules on CT image.The purpose of this paper is to construct a computer aided diagnosis system to identifybenign and malignant pulmonary nodules.This system is used to help the radiology doctorsmake early cancer diagnosis and reduce the repeated work of dr..High frequency noise whichwas introduced from CT machine are removed by using Gauss filter in the lung parenchymasegmentation.Images are processed into two values by Using adaptive threshold segmentationalgorithm.this way can achieve less manual intervention, improve the experimentalrepeatability, and eliminate the influence of the different threshold selection.using the localminimal algorithm to repair over segmentation of the image, it can avoid the computer imageedge curvature calculation process,and high efficiency and the pulmonary parenchyma partrelatively complete is guaranteed at the same time.A novel automatic feature assessment and weighting Fuzzy C-Means algorithm wasproposed for the classification in the paper.In the classification experiment between nodulesand other tissue in the lungs and between benign nodules and malignant nodules,thisclustering algorithm are used both.Pulmonary nodule extraction experiment clearly show thatautomatically weighted fuzzy C mean clustering algorithm classification more accurate fromthe picture.When this algorithm was used for the classification of solitary pulmonary nodules.The experimental results show that the accuracy of discrimination is86.3%, the sensitivenessis87.5%, and the specificity is80%, which illustrate that the method is feasible, and havegood accuracy and sensitivity.