Automated system for processing ultrasound images of the carotid artery based on evolutionary algorithms.

Abstract

This thesis is devoted to solving actual scientific task of improving the quality of automated segmentation of ultrasound medical imaging. The quality of automated segmentation of ultrasound imaging (UI) of the carotid arteries of man, specifically. The model of proposed flexible system is based on evolutionary algorithms. This model allows you to change the set of image processing algorithms which are used by the system for the synthesis of image segmentation schemes. This method of automatic synthesizing of UI processing schemes is based on the genetic programming. Image segmentation schemes are synthesized from the set of image processing algorithms. The thesis presents the method for reducing of a power of the set of image processing algorithms. This makes it possible to accelerate the synthesis of image processing schemes by 5%.

The developed system was tested with real sets of ultrasound images of the carotid arteries of 47 patients at different stages of the disease. The automating selection of image processing schemes is based on calculation of texture parameters. This makes it possible to increase processing accuracy of ultrasound images of human carotid arteries by 10-15% in comparison with peers.