Vessel Segmentation/Registration from Ultrasound Images

DESCRIPTION (provided by applicant): Ultrasound is widely used for imaging of blood vessels as it is non-invasive, real-time, and relatively inexpensive. This proposal focuses on segmentation of abdominal aortic aneurysms (AAA) from ultrasound images with extension to other vascular imaging applications in the long term. Reliable quantitative evaluation of AAAs plays a pivotal role in diagnoses and frequent follow-up studies needed to avoid life-threatening rupture. These studies require vessel segmentation (for size analysis) and registration between serial studies (for monitoring the progression of the disease before and/or after vascular repair). AAA evaluation is routinely carried out for both high-risk patient populations and those treated with endovascular repair. Currently, AAA management is primarily based on measurements from two-dimensional (2-D) slices in CT scans. AAA monitoring and follow-up could be improved by 1) measurement from 3-D reconstructions, and 2) use of ultrasound imaging to minimize radiation exposure and reduce costs. 3-D ultrasound reconstructions provide accuracy comparable to that of CT. However, large inter-observer variability and long processing times preclude routine clinical use of 3-D image information. This research aims to develop software solutions for improved ultrasound-based AAA monitoring and other vascular diseases (in the long term). The tools used will be based on advanced image segmentation and registration algorithms involving curvature-driven image processing techniques and deformable models. The goal of the Phase I study is to establish feasibility of the proposed methods by demonstrating an improvement in the repeatability and accuracy of measurements and reduction in delineation time.