High Precision Segmentation Of Pulmonary Airway Tree And Centerline Skeleton Extraction

In recent years, with the rapid development of medical imaging technology, multi-slice CT is widely used in clinical practice, so doctors can get a lot of high-precision CT slices which promote the related disease diagnosis. But the bulk of the CT slices produced the heavy task of reading them, which will bring the diagnostic doctors the problem of missing and false diagnosis. For this, to design a powerful computer-aided diagnosis system is the research focus currently.High-resolution CT scan of chest provided the possibility for the segmentation of the accurate airway tree and the diagnosis of lung disease, such as cystic fibrosis lung, emphysema, and trachea stenosis. So designing the lung airway tree automatic segmentation and disease automatic diagnostic system is an important branch of Computer Aided Diagnosis System.Affected by noise pollution and partial volume effect, the end of lung tube occurs the phenomenon of trachea fracture. With lung tube penetrates deeply into Lung parenchyma and trachea wall is gradually thinning, lung trachea will leak to pulmonary parenchyma. So it is prone to leakage in the process of segmenting lung airway. Overcoming the lung bronchial tube rupture and leakage, segmenting airway tree accurately is the key to construct a computer-aided diagnosis system, and is the basis for automatic diagnosis.On the base of segmentation of high-precision lung airway tree, the centerline of the lung airway tree is extracted, which is a single pixel wide, connected, smooth airway tree skeleton. It is an important guidance for the virtual endoscopic technique and anatomy-based markers for lung registration.The main research content and innovation as follows:Firstly, the medical image is segmented using ITK (Insight Toolkit), An United States released a C++ based Medical Image Segmentation and Registration of cross platform. We can research the segmentation algorithm of lung airway tree on the base of platform. The result of the segmentation is displayed by the method of surface rendering using the tool of VTK (Visualization Toolkit). Finally, a medical image segmentation and display platform is constructed by using ITK, VTK and MFC.Secondly, high precision segmentation of pulmonary airway tree can easily produce the problem of bronchial rupture and leakage. By studying and improving of the existing pulmonary airway tree segmentation algorithms, a method is presented in this paper based on optimal threshold region growth and morphology.Thirdly, on the base of lung airway tree segmentation, the skeleton of airway tree is extracted using the method of 3D images thinning algorithm. In the process of skeleton extraction, the position of middle line, the connectivity of objectives are ensured, and without the holes as well. Accurate extraction of the centerline of the trachea will have important guiding significance for Virtual Endoscope technology.Finally, the author especially thanks the National Natural and Science Foundation of China (60771007) and the Important Research Project from Anhui Provincial Bureau of Education (2006KJ097A) for funding on the related works.