Three-Dimensional Reconstruction of Cardiac Structures Using CT Images

Abstract

This paper describes a software tool for building and displaying tri-dimensional cardiac structures from a sequence of Computer Tomography (CT) scan images. The development process followed for this work can be divided into 5 stages. During the first stage, suitable DICOM images are selected from a CT scan machine. In the second stage, the images are pre-processed in order to reduce noise levels and enhance relevant features. The third stage compares the performance of three image segmentation methods on the pre-processed images: Region growing, Otsu's algorithm and active contours. It was found that active contours achieves the best isolation of the Region of Interest (ROI), while the other two methods recognized morphological structures outside of the ROI. In the four stage, a 3D image of the ROI is renderized. Finally, the reconstructed 3D model is visualized in stage five using a custombuilt Graphical User Interface (GUI). The resulting 3D models are validated by visual inspection by three heart and imaging experts, which determine if the model is suitable for use in diagnosis or surgery planning.