The Research On Automatic Segmentation Algorithm Of Lung Fields Based On Thoracic CT Sequence

In recent years, the environmental pollution problems have become increasinglyprominent, and hazy weather frequently appears in many countries and regions,resulting in respiratory system problems of many people. Lung cancer is the mostserious disease of respiratory system diseases, the incidence of lung cancer accountedfor the top of various malignant tumors. Each year there are about130million peopledied due to lung cancer around the world. If patients with lung cancer can be found,diagnosis and treatment in the early stage, the survival rate will be greatly increased.Therefore, early diagnosis and treatment of pulmonary disease is very important.Low dosage and multi-slice CT scanning is easy to find lung lesions and lesioncharacteristics, but with the development of CT technology, the number of slices oncedetected from a single patient’s lungs increases with the thickness of the slices to bethinner. If doctors confirm cases through view the images piece by piece, a largenumber of CT image data will bring the huge workload. With the wide application ofthree-dimensional visualization technology, more and more doctors are accustomed toanalyzing cases by the images’ three-dimensional reconstruction. However, CTexamination is to scan the patient’s whole chest, which also contains other organs andtissues, the other high-density tissue blocking the lung fields after the reconstructionof three-dimensional visualization, doctors can not view lung fields throughthree-dimensional reconstruction of thoracic CT images. Therefore, we need toremove the tissues and disturbed areas outside the lung fields to facilitate thediagnosis of the doctor.There are many types of lung disease and the imaging performance is extremelycomplex, in order to separate the lung fields with other organizations within the chest,and retain the afflicted region accurately improving the efficiency of the doctor’sdiagnosis, the automatic segmentation algorithm of complete lung fields from thoracicCT images was studied in this paper. Firstly, study the initial extraction method oflung fields. In the process of the initial extraction of lung fields, the method toremove the disturbed region of the adhesion of the trunk wall outside the trunk bymorphological or Euclidean distance transform and region filling was put forward.This method can remove the disturbed areas, while retain the initial outline of thelungs. Then analyzes several methods widely applied in lung region repair, including morphological closing operation, rolling ball and rolling circle method, a methodbased on local two-dimensional convex hull, and Euclidean distance transform torepair lung fields was put forward. The lungs boundary smoother after repair based onEuclidean distance transformation repair lung fields relative to other types of repairalgorithm, and repairing problems can be resolved when lung areas extracted fromthoracic areas are not only two. In order to remove trachea area better, finally, thispaper puts forward a set of automatic segmentation algorithm for lung fields based onthe three-dimensional morphological characteristics of the lungs. This algorithmintegrates the initial extraction and repair of lung fields, the lung CT sequence isdivided into four parts through the number of regional connectivity of trachea, thenumber of pixels of each regional connectivity of trachea and the circularity of thetrachea, which are the apex of lung, the middle portions of lung, the base of lungand the image does not contain the lung. By comparison with other commonly usedalgorithms, and the analysis of experimental data, proved that this automaticsegmentation algorithm for lung fields based on thoracic CT sequence has acertain clinical value.