Automatic coronary artery wall contour detection

L. Moura R. Kitney

The paper describes an image processing technique for the detection and labelling of arterial wall contours from cross-sectional slides in pathology. A gradient operator is applied to the digitised image and generates points of interest at likely edges. These points are then used as starting points for a contour follower algorithm leading to the formation of one-pixel-wide line segments. Heuristic search is used to loca te closed sets of segments. A cost-funtion is used which is minimised by a circle so that the algorithm tends to find roundish structures. Further processing involves closed curve filtering and con textual contour labelling. Results have shown that the method's strength lies in its ability to correctly find the wall contours even when major gaps are present, provided that some long segments are formed. The method is suitable for locating and labelling other round-shaped structures such as blood vessel cross-sections and cell outlines.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: