Vessel recognition in color Doppler ultrasound imaging

Abstract

Diagnostic ultrasound imaging is an important non-invasive tool for diagnosis and treatment decisions with a variety of clinical applications such as vascular disease, cardiology disease, and obstetrics. Color Doppler ultrasound imaging modality in particular offers complete 2-D cross sectional images depicting velocities of moving reflectors, such as blood flow. Color Doppler imaging is a crucial diagnostic tool for vascular disorders such as arterial atherosclerosis and deep vein thrombosis. Nowadays the vascular ultrasound exam is a tedious and time-consuming process that requires a high level of expertise. This is mainly due to the large number of manual controls offered by the ultrasound system to the clinical users to optimize the images for the best clinical outcome. Another reason is the lack of automation and quantification tools by the ultrasound system since the system utilizes very little high-level information from the acquired images. Moreover musculoskeletal injuries due to operating the ultrasound systems have become very common and a great concern for the medical community. Another concern is the standardization of the clinical exam outcome to minimize interobserver and intraobserver variability among diagnostic ultrasound practitioners. Introducing advanced automation and quantification capabilities to the ultrasound system will help with all these issues and streamline the clinical exam operation and outcome. In this thesis a complete vessel recognition system based on the analysis of the color Doppler ultrasound images is developed. The system includes the design and realization of many automated tasks including image acquisition, vessel segmentation, true velocity restoration, feature extraction, and classifier design. The involved tasks represent the building blocks for the automation and quantification applications that can be incorporated within the ultrasound system to achieve ease of use and more standardized clinical outcome. The thesis will also address a number of important clinical problems in the field of ultrasound vascular imaging and propose novel solutions that rely on the developed image understanding techniques.