Automatic detection of coronaries ostia in computed tomography angiography volume data

2016-10-05T05-46-18Z
Source: International Journal of Research in Medical Sciences
Noha Seada, Mostafa G. M. Mostafa.
Background: Heart coronaries emerge from the ascending aorta lateral sides from two points called the coronaries ostia. To automatically segment the heart coronaries; there must be a starting point (seed) for the segmentation. In this paper we present a fully automatic approach to segment the coronaries ostia towards automatic seeding for heart coronaries segmentation. Methods: Our algorithm takes as an input a CTA volume of segmented aorta cross sections that represents our region of interest. Then the ostia detection algorithm traverses that volume looking for the ostia points in an automatic fashion. The proposed algorithm depends on the anatomical features of the ostia. The main anatomic feature of the ostia is that it appears like a curvature or corner on the segmented ascending aorta cross section. Therefore we adopted in our methodology a modified version of Harris Corner Detection; besides inducing some anatomical features of the ostia location with respect to the aortic valve. Results: The proposed algorithm is tested and validated on the computed tomography angiography database provided by the Rotterdam coronary artery algorithm evaluation framework. The proposed automatic ostia detection algorithm succeeded to detect both ostia points in all the test cases. Also, the detected ostia points coordinates are validated versus a ground truth provided by the same framework with deviation between the results of the detection process and the ground truth having a min of 0 pixels and a max of 10 pixels for all test cases. Conclusions: Thus the proposed algorithm gives accurate results in comparison with the ground truth, which proves the efficiency of the proposed algorithm and its applicability to be extended as a seed for heart coronaries segmentation.


http://ift.tt/2cSUEB1