Development Of Computer-aided Diagnosis System For Pulmonary Function Based On Lung Tissue Segmentation And Detection Technology

The number of patients with various lung diseases such as lung cancer and chronic obstructive pulmonary disease has increased year by year,making accurate and efficient diagnosis of lung diseases gradually become the current research hotspot.Early pulmonary function tests usually use pulmonary function instrument which is based on mechanical and chemical methods,which is relatively cumbersome and time-consuming.With the rapid development of medical imaging technology,image processing technology and pattern recognition technology,the Computed-aided Diagnosis(CAD)system can not only provide quantitative indicators,but also perform three-dimensional visual assessment.Doctors have effectively reduced the burden of reading medical images and reduced the problems of missed diagnosis and misdiagnosis caused by repeated mechanical work.In order to assist doctors in the diagnosis of lung function efficiently and conveniently,this paper designs and develops a computer-aided diagnosis system for pulmonary function based on image segmentation technology,neural network detection of pulmonary nodules and threedimensional visualization technology.The research content of this paper is mainly divided into the following aspects:(1)In order to accurately extract the relevant lung tissue to accurately evaluate lung function,this paper studied the comprehensive segmentation algorithm of lung parenchyma,lung trachea and pulmonary vessels based on medical image segmentation technology.First,the preliminary segmentation of lung parenchyma is performed based on the combination of the largest inter-class difference method and region growing.Next,the segmentation of the lung trachea is completed by using the improved region growing method.Then,the pulmonary vessels are segmented by a method based on fixed threshold segmentation.Finally,the pulmonary parenchyma was repaired according to three steps: filling small holes based on morphological closed operation,filling large holes based on contour extraction,and repairing pulmonary edge defects based on convex defect detection.(2)In order to analyze the lung function of each lung lobe,a semi-automatic lung lobe method based on marker watershed algorithm is proposed.Firstly,the fissure is enhanced based on Hessian matrix,and each lobe area is manually labeled and expanded in three dimensions to form a lobe labeled image.Finally,the semi-automatic lobe segmentation is realized by combining the labeled watershed algorithm.(3)To assist doctors in observing the morphological characteristics of pulmonary nodules,judging their benignity and malignancy,confirming whether it will lead to abnormal pulmonary function,complementing and verifying each other with the pulmonary function indicators obtained,a pulmonary nodule detection method based on improved SSD neural network is proposed.Firstly,the application effect of traditional classifier in detecting pulmonary nodules is analyzed.Then,according to the characteristics of SSD network,the automatic detection of pulmonary nodules is realized by improving training data,changing the size of default box and adding deconvolution layer.(4)In order to obtain quantitative lung function indicators and visually display the distribution of lung tissue for visual assessment,a method for evaluating lung function based on 3D visualization technology and calculation of lung function index was proposed.The pulmonary function index was calculated based on the segmented lung tissues,and the comprehensive pulmonary function evaluation was realized by integrating three-dimensional visual evaluation.(5)Based on the above research content,a computer-aided diagnosis system for lung function is designed and developed independently in this paper.The system can segment the lung area and reconstruct its three-dimensional model,detect the pulmonary nodules,and calculate the relevant pulmonary function indicators of the whole lung,left lung,right lung and left and right lobes.It can realize the mixed rendering of multiple lung tissues and provide the interactive VOI(volume of interest)volume cutting function,so as to realize the quantitative analysis of local lung parenchyma,measure the pulmonary function indicators within the range of local lung parenchyma and interest CT value,and display them locally in two-dimensional and three-dimensional.The measurement and calculation of pulmonary function index and three-dimensional visualization function complement each other to realize the combination of quantitative analysis and visual evaluation.This system can be applied to the auxiliary diagnosis of chronic obstructive pulmonary disease and other lung diseases,and assist in evaluating the pulmonary function of patients.