Lung Cancer Detection And Resection Planning Based On Computed Tomography Images

Lung cancer is one of the most dangerous diseases with the highest morbidity and mortality in the world.It seriously threats human health.The current clinical study found that the early diagnosis and treatment of lung cancer could effectively prolong the survival of lung cancer patients and reduce the patient’s medical costs.There exist two main problems in the early diagnosis and treatment of lung cancer.The first one is the automatic detection of lung cancer,while the second is the intuitive and safe lung resection surgery.Therefore,this paper focuses on these two issues,and designs a complete lung cancer detection and resection planning system based on CT images.It can automatically and effectively detect lung nodules.Moreover,based on the visualization of the pulmonary vessels and the visualization of lung nodules,the lung cancer resection planning is more intuitive and safe.The paper focuses on the research and development of the system and mainly contains four parts:Stepped lung parenchyma segmentation based on CT images.Stepped lung parenchyma segmentation algorithm is proposed in this paper.The first step is to use automatic threshold segmentation and three-dimensional connectivity algorithm for the rough segmentation.The second step is to use the morphology and other methods,which fill the edge of the rough segmentation,for more refined segmentation.The validity of the segmentation algorithm and the integrity of the segmentation result are confirmed by qualitative and quantitative analysis.The parenchyma segmentation reduces lung cancer detection ranges and false-positive nodules,and decreases the computational complexity of the overall algorithm.Lung nodule detection based on two-step deep learning network.Firstly,fine annotation of lung nodules are generated based on conditional random field algorithm.Then,the improved U-Net network is used to extract the suspected regions of lung nodules.Then,the classification network is designed to remove the false-positive nodules in the suspected regions and to obtain the final lung nodule detection results.The validity of lung cancer detection algorithm is indicated by the FROC curve based on a large number of experimental data in this paper.Lung vessel enhancement and three-dimensional visualization.Lung vessels are enhanced based on the improved multi-scale Frangi algorithm.And then the enhanced value and gray value of the data is used to construct the two-dimensional transfer function of volume rendering.The threedimensional visualization of the enhanced lung vessels is implemented.Thus,it reduces misdiagnosis of blood vessels in the lungs and damage to other lung tissues during lung cancer resection.Computer-assisted lung cancer resection planning.Different lung tissues such as lung nodules,vessels and lung wall are assigned to different colors and opacities through classification visualization.Thus,the observation of the relative position of lung nodule and its surrounding tissues is more intuitive.Then,the best surgical path is adjusted by interactive simulation operations.It can assist clinicians for a more suitable surgical plan.Through the actual clinical data test,it confirms that the computer-assisted surgical resection of lung cancer improves the visual and safety of surgery in this paper.