Diameter Measurement of Vascular Structurs in Ultrasound Video Sequences

Abstract

Diameter measurements of vessel structures are of interest for a number of cardiovascular examinations. To support manual analysis of duplex ultrasound (US) images of human vessels and to improve the measurement accuracy and reproducibility, a semi-automatic approach has been developed which applies image processing techniques to compute reliably the vessel diameter. The proposed approach presents an interactive tool for measuring vessel diameters from US image sequences. A first derivative of a Gaussian (DoG) is applied for vessel wall detection followed up by a skeletonization step, and computation of horizontal line segments. Within a final classification step those line segments which show the highest likelihood are selected. The classification score is computed based on length, edge strength and linearity information. The overall approach has been evaluated at several sample sequences showing results outperforming“manual”measurements. It is currently applied in physiological studies.