Model-based segmentation of aortic ultrasound images

Morphological features of the aortic outflow ultrasound images are used in clinical practice for diagnosis of cardiovascular diseases. While feature extraction can be done manually, it is very time consuming. Segmentation is an important step in image interpretation, analysis, and quantification of the objects within a scene. In this work, we propose a novel method for the automatic segmentation of aortic outflow profiles based on a segmentation technique that incorporates a prior knowledge about the object shape in the form of the shape boundary model. The proposed model-based method utilizes a series of image analysis steps including image registration and a modification of the RANSAC algorithm to deal with noise and other artifacts in the image acquisition process. The experimental validation is done on a set of 67 patients and is compared to manual segmentation by an expert cardiologist. The proposed method has shown high correlation with results obtained by the expert cardiologist.