A method for extracting feature parts from CT-images of skull based on state transition model

A system for extracting characteristic bone regions, i.e., feature parts, from X-ray CT (computed tomography)-images (slices) of the human head is proposed. Such feature parts form regions of complex shapes and locations. Thus, to realize the system, a state transition model of the skull is first proposed, taking account of the topology of sliced skull shapes. Next, a sliced skull shape is recognized in each slice on the basis of the model. Finally, using the recognition results, feature parts are extracted from slices. An important feature of the method is the use of those search areas to guide the extraction process which are determined in each slice by constraints derived from the model and the recognition results. The method has been successfully applied to feature extraction of the skull