Model-driven contour extraction for physically deformed objects -application to analysis of stomach X-ray images

Proposes a model-driven method for extracting the contour of a physically deformed object using a flexible model which simulates the deformation. This method has been developed to extract a stomach contour from a double contrast X-ray image (abbreviated as the DC image). First, the method automatically constructs a stomach model based on the stomach shape extracted from a different type of X-ray image of the same stomach. The model consists of springs to represent the elasticity of a stomach. By using the model to simulate the stomach deformation between the two images, the stomach shape in the DC image is predicted. The predicted shape is registered in the DC image and is made closer to the exact contour by a two-staged method. Stomach contours were accurately obtained using 13 actual DC images