Electronic Thesis and Dissertation Repository

Degree

Doctor of Philosophy

Program

Biomedical Engineering

Supervisor

James C. Lacefield

Abstract

Operator-dependent instrument settings and the likelihood of image artifacts are two challenges for reliably using three-dimensional (3-D) power Doppler angiography in flow depiction and quantification applications.

To address the operator-dependent settings challenge, an automated method for wall filter cut-off selection, the wall filter selection curve (WFSC) method, was developed using flow-phantom images. The flow-phantom WFSCs guided the development of a theoretical signal model relating color pixel density (CPD) and wall filter cut-off frequency. Simulations using the theoretical model were used to define criteria for the WFSC method to be applied to unprocessed power Doppler signals from 3-D vasculature. The adapted WFSC method was combined with a 3-D skeletonization and vessel network reconstruction method to present a two-stage processing method aimed at improving vascular detection, visualization and quantification. The two-stage method was evaluated using two in vivo models; a murine tumor model was used to test the performance of the method in a flow quantification application and a chick embryo chorioallantoic membrane (CAM) model was used to evaluate the method’s value for flow depiction applications.

Applying the WFSC method to flow-phantom images improved vessel delineation and vascular quantification to within 3% of the vascular volume fraction of the phantom. Criteria for the WFSC method from the simulations were to assess at least 100 cut-off frequencies and that the CPD variability should be less than 5% to ensure quantification accuracy. Large variations in the cut-off frequency selected using the WFSC among images acquired at different time points and across different animals in the murine tumor model signified the relevance of spatially and temporally adjusting the cut-off frequency. The two-stage method improved visualization of the vascular network and significantly reduced artifacts in both the tumor and CAM models in comparison to images using conventional Doppler processing. In the CAM model, vessel diameters measured in two-stage processed images were more accurate than measurements in images exported from a commercial scanner. The proposed signal processing methods increase accuracy and robustness of qualitative and quantitative studies using 3-D power Doppler angiography to assess vascular networks for flow depiction and quantification.