Aproksimacijos elipse taikymas galvos kompiuterinių tomogramų analizėjė

Abstract

In the report there are reviewed modern ellipse fitting algorithms and its application in the analysis of human brain computed tomography images. Special objects of human brain tomography images are head bones which are easily detectable in every slice as bright ring-shaped areas. It is able to approximate the bone region by using ellipse fitting algorithms based on optimization and clustering (voting) methods. In the result geometrical features of head bone as well as of brain area may be evaluated, for instance, symmetry axis, centre, eccentricity. Ellipse fitting is also useful in identification of brain region because ellipse contour can separate brain from other tissues. Both optimization and clustering algorithms are suitable for analysis of computed tomography images if used together with special preprocessing. Experimental results show that angle between real and modelled symmetry axes does not exceed 10 degrees in more than 60% of processed images and brain area was detected without deviations in more than 40% of analyzed slices.