Recognition of abdominal organs using 3D mathematical morphology

This paper describes a method for recognizing abdominal organs in a computed tomography image primarily based on mathematical morphology. We first identify the coordinate origin and the scale in the abdominal region and then set up four regions for recognition. The recognition process consists of two steps: segmentation and identification. For the segmentation step we employ a differential top-hat (DTT) operation. For the identification step, we first identify the vertebra that the lowest rib meets. The gravity center of this vertebra is used to set the coordinate origin. In three regions, we separate organs based on threshold values, while in a region containing the liver and stomach we use RE (Recursive Erosion) and GI (Geodesic Influence) in order to separate touched organs. The separated organs are identified or labeled based on the size and position. This method is judged to be effective since we were able to obtain a recognition or labeling rate of approximately 91%. © 2002 Wiley Periodicals, Inc. Syst Comp Jpn, 33(8): 75–83, 2002; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/scj.1148