Research On Key Technologies Of Computer Aided Teaching For CT-based Lung Cancer Diagnosis

Effective disease diagnosis from medical images is an important clinical practice skill for students of imaging diagnostics.By studying the medical image characteristics of the body in a normal or disease state,learners can establish the pathological process of the diseases and the relationships among anatomical structures of the human body,so as to effectively guide the clinical treatment of medicine and surgery.However,due to the influence of instrument parameters or external environment,the image quality is often not high,and the visualization of teaching content and the intellectualization of examination and evaluation are insufficient,which cannot meet the teaching needs.Abstract teaching content and lagging teaching methods are not conducive to developing teaching activities and hinder the construction of learners’ clinical thinking and the effective diagnosis of diseases.With the development of the Educational Informatization 2.0 Action Plan,China’s education has entered a new era of informatization.The deep integration of information technology and education is helpful to promote the teaching reform of intelligent medicine.Based on the actual demand of CT diagnosis teaching of lung cancer,this paper studies the related problems and key technologies in CT diagnosis teaching of lung cancer.The main work and innovations of this paper include:First,the factors causing low visualization degree of CT images such as noises are analyzed to enhance the detectability and recognition effect of detail information in CT images,and the CT image preprocessing technologies are studied including a smoothing method based on empirical mode decomposition and bilateral filtering,a denoising approach based on normal vector similarity and non-local mean filtering,and an enhancement technique based on multi-scale dark channels.The smoothing method can help enhance the strong edge of lung tissues and eliminate the noise and weak edges in CT images.The denoising method can remove the noise in CT images.The enhancement method can enhance the detail representation of CT images effectively.The preprocessing technologies proposed in this paper are helpful for learners to observe the details of lung tissues,as well as the subsequent segmentation of lung tissues and feature extraction of lesions.Second,in order to facilitate the display of specific teaching content,quantitative assessment of lesions and student assessment,and to provide accurate lung tissue regions for three-dimensional visualization,the lung tissue segmentation technologies are studied in this paper.Firstly,lung region segmentation methods based on random forest,deep convolutional neural network,and the combination of random forest and deep convolutional neural network are proposed,respectively.A series of post-processing approaches are designed after the extraction of initial lung regions with the above methods to further improve the technique accuracy.Secondly,the methods of lung nodule extraction based on multi-scale edge detection and grey correlation are presented,respectively.The methods provide efficient and accurate segmentation results,which offers important support for lung tissue visualization and intelligent evaluation.Third,semantic characteristics of pulmonary nodules are helpful for intelligent diagnosis of lung cancer in computer-aided teaching.The semantic characteristics of lung nodules are introduced in this paper,and the correlation between the semantic features(spiculation,lobulation,calcification,subtlety,margin,sphericity,internal structure)and malignant degree of pulmonary nodules are analyzed.Classification techniques based on deep neural networks are proposed to grade the semantic characteristics with high correlation with the malignant degree of pulmonary nodules.Semantic characteristic grading can effectively assist the classification and the malignancy evaluation of nodules in the intelligent diagnosis of lung cancer.Finally,to present the lung tissue information intuitively,truly,and comprehensively and assist medical students to have a deep understanding of lung tissue and lesions,the sparse CT image 3D reconstruction performing in 3D point cloud space is studied based on the analysis of the existing 3D visualization model reconstruction techniques.First,lung tissue contours in CT sequence images are extracted by a single-pixel tracking algorithm,then the contours of missing lung tissues are constructed by the GVF Snake model,and point clouds and corresponding normal vectors are calculated.Finally,lung tissues are reconstructed by reconstruction algorithms,such as the floating scale algorithm.The reconstruction technique can effectively deal with the problems of matching and branching in the reconstruction of lung tissues and reconstruct accurate lung tissues.The research on the key technologies of computer aided CT diagnosis teaching for lung cancer is helpful to enhance the learning experience of learners,solve the teaching difficulties of teachers,and effectively assist the teaching activities of medical imaging diagnosis,so as to promote the deep integration of artificial intelligence technology and medical teaching and promote the development of medical education informatization.