Chest HRCT Image Segmentation Based On Granularity Synthesis

In recent years, many researches have indicated that a variety of lung diseases are related to the pathological changes of small airways. Therefore, the diagnosis of small airway diseases is very important. High-resolution CT(HRCT) is one of the important imaging means for the diagnosis of the small airway diseases, it can reflect the lung's overall condition, and can distinguish small lesions, thereby it can provide doctors with more accurate and detailed information in the diagnosis of small airway diseases. However, the features of wide range of gray level distribution and Complex textures of the image result in the difficulty of image feature analysis and tissue segmentation. At the same time massive HRCT datas increase doctors'burden, in order to improve the efficiency of the doctors'diagnosis and reduce their labor intensity, computer-aided diagnosis (CAD) system came into being. When using CAD to help doctors diagnose small airway disease, the one problem that must be solved is the accurate segmentation of lung tissue. In this dissertation, we do some researches for the lung region's exact segmentation. The main work is as follows:(1) The features of HRCT image are described, HRCT signs of several typical small airway diseases are analyzed, the MATLAB image processing toolbox is studied, which is used to process the images with DICOM format.(2) The defect of the conventional clustering segmentation method is that the clustering centers are appoint artificially, therefore the segmentation result is inaccurate. In this paper, a segmentation method based on the HRCT images gray feature is proposed, in this method the clustering centers are appointed according to the extracted graylevel histogram of the HRCT image. We also do the segmentation based on the texture feature, use wavelet transform method to extract texture feature of HRCT images, and complete the image segmentation.(3) As a result of the segmentation results based on a single image feature are not satisfactory, we propose a HRCT image segmentation algorithm based on granular synthesis. Firstly, the granularity theory contained in HRCT image segmentation is elaborated; secondly, the granularity model is established; finaly, the texture feature and gray feature are combined by granular synthesis method to achieve the image segmentation. Experiments show that the proposed method can effectively use the underlying features of the images, and has a higher segmentation accuracy than traditional segmentation method, and lays the foundations for the medical image processing and analysis.