Tracing Based Segmentation for the Labeling of Individual Rib Structures in Chest CT Volume Data

摘要

We propose a fast and robust segmentation algorithm for the extraction and labeling of individual rib structures in chest CT volume data. A diagnostic system based on this algorithm can display 3D rib centerlines, contours and surfaces. A click on a rib point in any slice image will have the system identify instantly the individual rib it belongs to. The algorithm is based on a recursive tracing approach. The geometrical properties of the rib structure are explored to set up valid assumptions and model. At each step, statistical analysis is combined with dynamic programming to estimate the outer surface contour from the detected edges. The algorithm works reliably on CT volume data of variant doses and resolutions. This algorithm can be extended to other modalities. The detected centerlines can also be used for reliable and fast registration of in or cross-modality volume data of chest scans.