6 Summary and Future Work 157
6.1 Summary 157
Vascular imaging and quantification, typically called angiography, is becoming a core need for almost all clinical and research application involving pathology diagnosis, treatment and surgical interventions. Ultrasound stands its position as a valuable vascular imaging technique with its low cost, lack of ionizing radiation and being the non- invasive. Doppler ultrasound flow imaging techniques provide a wealth of information about vascular characteristics that can be presented using a number of different display modes. Power Doppler is characterized by its ability to image small vessels with slow flow, which makes it particularly useful as a vascular quantification tool. However, the sensitivity to operator-dependent instrument settings and the likelihood of image artifacts are challenges for quantitative power Doppler imaging. Therefore, development of new signal processing methods to overcome some of these challenges can enhance the usability of power Doppler imaging in quantitative microvascular angiography.
6.1.1 Chapter 2: Improved objective selection of power Doppler wall-filter
cut-off velocity for accurate vascular quantification
An improved (relative to the original method presented in[1]) automatic, objective method for the selection of one of the ultrasound operator-depend instrument settings, namely, the wall filter cut-off frequency is presented. The method, called the wall filter selection curve (WFSC) method has three key features: first, it automatically identifies the cut-off frequency ranges enclosing the optimum cut-off frequency based on the characteristics of the imaged vasculature, secondly, using a multi-step decision algorithm, the method identifies an operating point within these ranges. Finally, using the first and second features, the method performs spatial tuning of the cut-off frequency by selecting a cut-off frequency for subregions within a region of interest. Evaluation the WFSC method using multiple-vessel flow phantom images demonstrated the improved depiction
of blood-mimicking fluid flow with smoother and sharper vessel boundaries in comparison to images processed and displayed using typical commercial scanner software. Vascular quantification of images processed using the improved WFSC method was accurate to within 3% of the vascular volume fraction of the phantom.
6.1.2 Chapter 3: A new three-component signal model to objectively select
power Doppler wall filter cut-off velocity for quantitative
microvascular imaging
Building on the theoretical basis of the WFSC method presented in Chapter 2 and using experimental WFSCs from flow-phantom images, a new signal model describing the relationship between wall filter cut-off frequency and the ratio of colored pixels displayed in a power Doppler image (color pixel density (CPD)) was developed. The new model showed an improved fit to experimental flow-phantom selection curves in comparison to the original theoretical model developed in[1]. Monte Carlo simulations of different vascular environments by changing the number of vessels and the mean and standard deviation of blood and background tissue velocity distributions in conjunction with reference values derived using a cost function were used to analyze the performance of the WFSC method to automatically select an operating cut-off frequency. These simulations were also used to identify conditions necessary for the development of an online implementation of the WFSC method including the reliable number of samples on a selection curve (100 samples) and the upper limit to fluctuations to CPD within a region of interest (5% of reference vascular volume fraction), the length of a reliable characteristic interval (longer than 6mm/s) and the threshold for the starting cut-off velocity of an interval is set to 3 mm/s. Evaluation of the accuracy of cut-off selection using reference values from the cost function supported the satisfactory performance of the multi-step decision algorithm for operating cut-off frequency selection developed in Chapter 2. This theoretical analysis was necessary to establish the expectation that the WFSC method can improve the accuracy and reproducibility of power Doppler for quantitative microvascular imaging by adapting the cut-off frequency to spatial and temporal variations in blood conditions.
6.1.3 Chapter 4: A two-stage process to improve quantitative three-
dimensional power Doppler angiography of tumor
microvasculature
A three-dimensional vascular network reconstruction method was developed and combined with the improved WFSC method from Chapters 2 and 3 to present a two-stage power Doppler processing method targeted at improving the quantitative performance of power Doppler angiography. Evaluated using power Doppler images of a murine tumor model, the two-stage method showed improved visualization of the vascular network. Quantifying vasculature using power Doppler angiography indices (vascularization index, flow index and vascularization flow index (VFI)) showed significant variations when images processed using the two-stage method were compared to images processed using typical, fixed wall filter cut-off frequency without vascular network reconstruction. Small improvements in correlation of the power Doppler angiography metric, VFI, with contrast-enhanced ultrasound image quantification over typical Doppler-processing images were reported. The first in vivo evaluation of the performance of the improved WFSC method (Chapter 2) demonstrated the relevance of spatially and temporally adjusting the cut-off frequency within a 3-D image as the mean WFSC-selected cut-off showed large variation within each 3-D image, and among images acquired at different time points across different animals (two-way ANOVA, p < 0.0001). The vascular feature responsible for these variations of cut-off frequency selection was identified as the ratio between total vascular length and vascular volume. These results suggest that the two-stage process has the potential to improve the reliability of visualization and quantification of complex, dense vasculature using 3-D power Doppler angiography.
6.1.4 Chapter 5: Improving microvascular depiction in three-dimensional
power Doppler ultrasound using a two-stage processing method
The two-stage method developed in Chapter 4 was evaluated using an in vivo
model of the chicken embryo chorioallantoic membrane as a vascular depiction application of power Doppler imaging. Applying the two-stage method to power Doppler signal data improved vessel detection and visualization and resulted in significant image artifact reduction in comparison to images processed using a commercial scanner software and set up by a licensed sonographer. Vessel diameter measurements from
images processed using the two-stage method were more accurate than measurements made using raw images exported from the commercial scanner (with a median absolute percentage error = 10.39% versus 28.18% respectively). Further analysis of the types and orders of vessels contributing to the overall error showed that higher errors in diameter measurements were associated with vessels of lower orders (orders 6 and 7, which have diameters of approximately 280 to 550 µm). It was also shown that a diameter
overestimation bias of approximately 30% was observed for arterial diameter measurements using scanner exported images whereas measurement errors from the two- stage method images were centered at 0%. Thus, the proposed method shows promising results for improving vascular quantification, detection and visualization using 3-D power Doppler imaging suitable for flow depiction applications.