3D Visualization And Computer Aided Diagnosis Based On Heart Images

Computed tomography images and three-dimensional echocardiography have important clinical implications for diagnosis and surgical planning of a variety of adult cardiovascular disease, and congenital heart disease; As an important clinic application of 3D visualization, cardiovascular virtual endoscopy has become a hotpot of computer aided diagnosis. This paper studies the 3D visualization technique and computer-aided diagnosis technique based on these two images. Based on Computed tomography, we study 3D visualization technique containing path planning, navigation design and rendering acceleration of cardiovascular virtual endoscope and the aided diagnosis taken on the basis of visualization; Based on the three-dimensional echocardiography we focus on finding the standard four- chamber view collected from the apex of the three-dimensional echocardiography in our research work. The major contributions of this paper are as follows:1. After reading related research literatures on path planning algorithms of virtual endoscopy widely, we summarize and compare common path planning algorithms according to the characteristics and focus on the principles, applications, advantages and disadvantages of centerline extraction algorithm based on distance mapping. All works above laid a foundation of the application of distance mapping algorithms in our cardiovascular virtual endoscope.2. An improved centerline extraction algorithm based on distance field is proposed for the cardiovascular virtual endoscope. The proposed algorithm is based on the traditional distance field method and avoid the hug-corner problem。Based on the automatic method we put forward an interactive which makes the cardiologists observe the lesions more closely and clearly.3. We use CUDA to accelerate the ray-casting volume rendering algorithm which greatly speeds up the movement of the virtual viewpoint along the planning path. Meanwhile we make the navigation design for our planning path that guarantees the viewpoint will not be lost during the endoscopic diagnostic progress. Based on the above work we do lots of experiments on patients with congenital heart disease, displaying the malformation we can diagnose.4. A method based on image retrieval is put forward to looking for the standard four-chamber view automatically in three-dimensional echocardiography. The method contains two part, in part one we get all four-chamber views using SVM classifier and GIST feature and in part two we use a template view to acquire the standard four-chamber view based on PCA.