Research Of Key Algorithm Of Lung Nodule Detection Based On Multi-projection Images

With the development of computer science technology, computer-aided diagnosis (CAD)has been widely used in medical imaging and diagnosis. The lung nodule detection is a typicalapplication of CAD. Multi-projection correlation image has been used for lung noduledetection in clinics for its simple equipment and low dose. However, the complexity of tissuesand the limitation of imaging system lead to a fuzzy edge and a poor contrast of themulti-projection correlation image, and these factors will directly affect the performance oflung nodule detection and lung field segmentation. So the research of key algorithm has avery important application value for lung nodule detection.In this thesis, we first make the analysis and summary of lung nodule detection methodin existence, and then in-depth study the theory of lung nodule detection and lungsegmentation. We purpose the developed algorithm of lung segmentation, and design amethod of lung nodule detection.Firstly, a lung field segmentation method using feature images and gray and shape modelis proposed.6feature images are calculated and an initial shape model is built. The dynamicprograming is used to search the optimal lung profile based on the gray cost and shape costwith the feature images. The lung profile is modified using Active Shape Model. Theexperiment results indicate that the method proposed in our study can improve theperformance of lung field segmentation.Then we have implemented the key algorithm of lung nodule detection. Thesegmentation method of feature images and gray and shape model is used. The method ofmulti-threshold segmentation is used to detect the candidate nodules and the dynamicprogramming is used to segment the nodules accurately. We use liner classifier to classify thelung nodules based on the features which have been extracted. The nodules which aredetected in one case are registered according to the correlation of multi-projection correlationimages, and the false positive nodules are removed.