Research On Automatic Segmentation Of Lung Trachea CT Images Based On Three-dimensional Region Growth Method

With the rapid development of medical imaging technology,CT(Computed Tomography)images have become an important basis for diagnosing diseases.In modern society where respiratory diseases are high,if traditional diagnostic methods are used,doctors face a single case of hundreds of images of lung disease,and the brain imagines the exact shape and overall contour of the trachea and blood vessels on each image.The diagnosis method not only increases the workload of the doctor,but also relies heavily on the doctor's personal experience in the diagnosis of the disease,and the diagnosis result of the disease is easily biased by the doctor's subjective consciousness.With the application of computer technology in medical images,it is urgent to use computer technology instead of manually extracting regions of interest from a large number of case image data,and generating intuitive three-dimensional information,so that doctors can directly observe the space in the region.Location,current shape and size.However,chest CT images are not only complex in structure,but also sensitive to noise,and there are large individual differences in CT values and morphological structures between lung trachea and pulmonary vessels,making segmentation of pulmonary tracheal trees and pulmonary vascular trees particularly complex.In view of the above problems and difficulties,this paper studies the preprocessing algorithms of chest CT images,lung tracheal primary segmentation and fine segmentation algorithms,and pulmonary vessel segmentation algorithms.Firstly,In the aspect of chest CT image preprocessing,this paper adopts a highly efficient and simple chest CT image preprocessing scheme.Through the morphological processing and smoothing operation,the influence of noise on the region of interest in the chest CT image is reduced,and then the hole in thelung image sequence is repaired by the flooding filling method,so that the segmented lung region contains the intact lung texture.Secondly,In the aspect of lung tracheal segmentation,when the lung trachea is initially segmented,the regional trough tree based on the hysteresis threshold is used to segment the lung tracheal tree at the optimal threshold.When the tracheal tube is finely segmented,this paper combines the tubular structure characteristics of the lung trachea in three-dimensional space.Based on the similarity between local pixels,a lung tracheal tree segmentation algorithm based on region growing and Hessian matrix enhancement is proposed.Finally,In the design and implementation of the three-dimensional lung texture assisted diagnosis system,using C++ as the programming language,combined with the open source VTK and ITK open source framework,the automatic segmentation and 3D visualization of the lung region and lung texture are realized.Completed a set of auxiliary diagnostic systems that can be used to provide a realistic and objective 3D visualization of the doctor.