Arterial Extraction Method In CT Images

Accurate segmentation and extraction of arteries in CT images with image processing technology,helps the medical research workers better engaging in research and treatment of related diseases.Due to the uneven distribution of brightness of arteries and their branches in CT images,and the interfering of brightness similar tissues to the arterial seed point location,the existed arterial segmentation and extraction methods are not satisfactory till now,nor the real-time needs in practical application are satisfied.a new automatic extraction method is proposed for the arteries in CT images from chest to hip.Firstly,this method locates the arterial seed points automatically,then segments spine bone and associated tissue area which may interfere the extraction of arteries,and then extracts the arteries using region growing,finally it improves the arterial extraction results through morphological repair and the extraction of small branches.Aiming at the automaticity of artery extraction,a method to position seed points automatically and accurately according to the arterial trunk characteristic is put forward.For the problem that spine bone interfers the arterial extraction,a method to segment the region of spine bone and the associated tissues by using the strategy of layer partitioning and OR operation combination for multilayer images is proposed.To overcome the small branches missing problem in the extraction,the idea of set and hierarchical region growing are used to obtain satisfactory results.The arteries extracted by this method are the parts of the descending aorta from chest to hip.The effectiveness of the proposed method is verified by testing 200 sets of data obtained from CT imaging devices from different manufacturers.Related medical personnel think that basically the extracted arterial data meet the requirements of clinical uses through evaluating them.The test data are three-dimensional images,and the size of each single image is 512 x 512 pixels,while the number of image layers is within the range [180,850].For each test example,the processing time is roughly from 5 to 20 seconds,which meets the requirement of real-time analysis.The method has been composed into a product by the cooperative research company and put into use.