3D Visualization Of Cardiovascular CT Images & Computer Aided Diagnosis Techniques

With the increasing pressure of modern life,cardiovascular disease has become the biggest threat to human health.Currently,the clinical diagnosis of cardiovascular disease mainly depends on CT angiography,echocardiography,magnetic resonance imaging and other medical imaging techniques.Medical images contain extremely rich information,but due to the specificity and particularity of its imaging modality and imaging object,medical image is usually not easy and intuitive to understand.Thus,3D visualization of medical images and other computeraided diagnosis techniques combining with computer vision,computer graphics and pattern recognition etc.theory are developed accordingly.This paper mainly studies the 3D visualization of cardiovascular CT images and its ancillary diagnostic technique in clinical applications.In this paper,we use cardiovascular CT images as research object,focusing on the cardiovascular CT images 3D visualization and its application on computer-aided diagnosis.By widely reading relevant literature and research achievements of predecessors,our group implements a virtual endoscopic cardiovascular system,this paper conducts a research and analysis on it and summarizes the Cardiovascular Virtual Endoscopy into five key components,image preprocessing cardiovascular system,three-dimensional reconstruction,cardiac segmentation and extraction,roaming path planning and real-time display of the endoscope.In 3D visualization,this paper focuses on the volume rendering technique,especially makes deep study on 3D texture mapping and ray casting method,and makes a comparison between 3D texture mapping and CUDA-based ray casting method in 3D reconstruction.For virtual endoscopy roaming path planning,we introduce a improved method compared with the original method.For distance transform,we introduce a general linear time method and use CUDA to do the parallel optimization;for Dijkstra centerline extraction,we use an advanced data structure called pairing heap to improve Dijkstra algorithm.Subsequently,we design an automatic coronary stenosis detection method based on the improved centerline extraction method by segmenting and 3D reconstructing coronary artery,then extract the centerline and calculate the area of coronary to calculate the extent of coronary artery stenosis.The experiment results show that the improved roaming path planning method in this paper has better time efficiency and the coronary artery stenosis detection method designed in this paper also has a good performance.The automatic detection of coronary artery stenosis methods proposed herein for cardiovascular research and the key steps of the virtual endoscope system optimization and this design manual for current clinical diagnosis of cardiovascular disease have a good secondary meaning.The research and optimization on key techniques of cardiovascular endoscopy system and the coronary artery stenosis detection method proposed in this paper can both provide good assistances for clinical diagnosis in cardiovascular disease.