Research On Nodule Detection Method In Lung CT Images Based On Three-dimensional Reconstruction

Lung cancer is currently the leading cause of cancer-related deaths worldwide.Most lung cancers are not diagnosed until the cancer has metastasized,because symptoms usually do not appear until the cancer is in an advanced stage.By improving the early detection rate of lung cancer,early treatment,can improve the survival rate of patients.After the three-dimensional reconstruction of lung CT,the contrast between the nodules and the background can be effectively improved,and the detection rate of nodules can be improved.Meanwhile,it can make the doctor observe the nodules more visually and visually,and improve the diagnostic efficiency and quality of doctors.Therefore,it is very important to study the three-dimensional reconstruction of lung CT and the detection of pulmonary nodules.Because of the small size of the lesions and the small inter-class differences in medical images,pulmonary nodules usually account for only a small part of the CT image.Firstly,the feature of this small region may not provide enough information due to its spatial feature resolution.Secondly,in two-dimensional CT scan,it is often difficult to distinguish between the lesion area and the non-lesion area because they may have very similar characteristics.In order to solve these two problems,it is necessary to enrich the expression of lung CT features,so as to improve the accuracy of lung nodules detection.The effectiveness of lung nodule screening depends to a large extent on the skill and experience level of the radiologist.Due to the complex structure of lung tissues,potential malignant lesions are easily overlooked.In order to solve the above problems,this paper proposes a 3d reconstruction based lung CT image nodule detection method.The proposed lung nodules detection method can obtain the spatial features of the images through the three-dimensional reconstruction of the CT images of the lungs,which can enhance the differentiation between lesions and non-lesions and improve the detection accuracy of lung nodules.In this paper,the 3d reconstruction of pulmonary CT images and the detection of pulmonary nodules are divided into two stages.The first stage is the 3d reconstruction of pulmonary CT images.The second stage is to study the detection method of three-dimensional pulmonary nodules.Through the study of these two stages,the goal of improving the detection accuracy of pulmonary CT nodules is finally realized.Therefore,this paper mainly studied the multi-resolution reconstruction method of lung CT images and the detection method of three-dimensional pulmonary nodules.Aimed at the first stage,because lung nodule in lung CT image of small,the bronchial lung,blood vessels,such as organization and are similar in two-dimensional CT features of pulmonary nodules,interference of nodule detection,aiming at this problem,this paper on enhanced CT images,the de-noising pretreatment,and extracting ROI regions,namely nodule area,and then for ROI area high resolution 3 d reconstruction,to low resolution 3 d reconstruction of background region,improve the ROI region and background contrast,both got high resolution 3 d lung nodules are visible,and can retain nodules in CT image 3 d background information,To achieve more accurate detection of pulmonary nodules from CT images.The proposed multi-resolution 3d reconstruction method is faster than the traditional method and can improve the resolution of pulmonary nodules and surrounding tissues more effectively.For the second phase,because on the two-dimensional slices,blood vessels,such as bronchial organization and nodule characteristics highly similar,there are interference of nodules detection,this paper presents a three-dimensional detection of lung nodules CT images,is the first phase of the reconstruction of 3 d lung CT images,input YOLO-3 d network framework,to extract the three-dimensional space characteristics of pulmonary nodules,and then to detect lung nodules,final output nodes in 3 d space coordinates.In the experimental results,the average accuracy and sensitivity of the method in this paper reached 73.67% and 95.83% respectively,which proved that the method could obtain a higher detection accuracy of pulmonary nodules.