- Top of page
Quantitative three-dimensional (3D) power Doppler angiography represents the acquisition and measurement of power Doppler data within a 3D dataset1. Increasingly, this technique is being used to compare different patient populations and changes within the same population2. The majority of these studies use the ‘histogram’ facility (GE Medical Systems, Zipf, Austria), which displays the distribution of the power Doppler data and uses specific algorithms to derive three indices of blood flow within a volume of interest (VOI) defined by the user3. The vascularization index (VI) represents the relative proportion of color voxels to gray voxels, the flow index (FI) represents the mean power Doppler signal intensity value, and the vascularization flow index (VFI) is a combination of the two, calculated by multiplying the values together and dividing the result by 1004.
These 3D vascular indices depend on, and relate to, the total and relative amounts of power Doppler information within the VOI and the intensity of the signals5. The power Doppler signal is dependent on the presence of blood flow within the tissue or organ being studied and its intensity is dependent on the number of scatterers or blood cells within the blood vessels. The intensity of the power Doppler spectrum is determined by several settings, all of which are designed to induce an increase or decrease in the signal, as required6. It is likely that these settings affect the 3D vascular indices, but the precise effects and their extent have yet to be determined.
Whilst many authors are aware of the importance of maintaining Doppler settings between subjects in the research setting, there are no published studies to confirm this or to evaluate the magnitude of any effect produced by changing settings. The objective of this study, therefore, was to use an ‘ultrasound flow phantom’, a Doppler test device that can be used to mimic the in-vivo setting and provide a stable environment in which to examine the effects of machine settings7, to determine how the 3D vascular indices generated by the histogram facility are affected by different power Doppler settings.
- Top of page
This is the first in-vitro phantom study to quantify the effect of different Doppler settings on the 3D vascular indices generated through the application of the histogram technique to quantify power Doppler data. There was a quantifiable effect for the majority of changes in any of the settings, highlighting the importance of strict maintenance of these settings for research and within the clinical setting if any two subjects are to be compared. These findings can be explained by considering the derivation of the power Doppler signal and how it can be manipulated and displayed for clinical use.
The majority of Doppler settings are adjustable to allow the observer to maximize the power Doppler display, dependent on the depth and flow characteristics of the area of interest. Any change in the settings will induce an increase or decrease in the Doppler signal, therefore, and, in agreement with theoretical predictions based on the physical properties of power Doppler and results from other in-vitro experiments, we observed an increase in the 3D vascular indices with increases in power and color gain, and decreases in these indices with increases in the PRF and wall motion filter. There was also an increase in the vascular indices when the signal rise was reduced and when the persistence of the signal was increased. This might be expected, as these values will lengthen the time that the signal is on-screen and the time of its capture during the 3D sweep.
While the integrated power spectrum displayed using power Doppler is proportionate to the total number of scatterers and their mean speed, the nature of this relationship is also affected by several other factors9. The autocorrelation function of power Doppler is proportionate not only to the physical properties of the BMF but also to the properties of the Doppler imaging system10. The findings in this experiment were expected, therefore, but the relationship between each of the three different 3D vascular indices and the various Doppler settings was more complex than that predicted. A fairly linear relationship was seen between VI and VFI and power, gain and wall motion filter, while the relationship was more curvilinear between FI and these indices, with a trend towards a plateau at higher FI values. PRF was associated with complex changes in all three indices, although this probably reflected the changes in power and wall motion filter that occurred automatically with adjustment of the PRF and were not modifiable. Both increases in power that occurred over the range of PRF values examined led to an increase in the vascular indices to values above those calculated at the two previous lower PRF values, suggesting that power outweighs the effect of PRF to some degree. The relationship between signal rise and persistence was also complex and more difficult to account for. VI was relatively stable over the first several increments in signal rise before falling suddenly at a value of 0.8, while, with an increase in signal persistence, it increased more linearly, albeit somewhat erratically. In contrast to the other parameters, FI had a more consistently linear relationship with both variables.
Jain et al.6 also demonstrated a negative effect of PRF and wall motion filter on the strength of the power Doppler signal, using an in-vitro left heart pulse duplicator system. The flow area derived from the power Doppler spectrum was reduced by successive increases in PRF and color filter, while it increased with a decrease in the frame rate. Yoon et al.7 used microspheres as BMF to examine the effect of flow velocity, wall filter, PRF and gain on power Doppler signal intensity and background noise, using computerized quantitative analysis. In keeping with our results, they reported that the intensity of both the flow signal and background noise was proportionate to flow velocity and gain but inversely proportionate to PRF and wall motion filter level. These parameters also appear to have an important influence on background noise and the potential for artifacts, as increases in both PRF and Doppler gain produced a high signal-to-noise ratio when the wall filter was kept constant, as did increments in the wall filter and gain at a constant PRF. This relationship was reversed when a high constant Doppler gain was maintained and the wall filter and PRF were reduced. Gudmundsson et al.11 also showed a significant effect of several instrument settings on the power Doppler display using offline analysis of signals derived following the digital conversion of data captured from video recordings of flow within a test device The most important variable was fluid flow velocity, although depth and angle of insonation were also found to be important determinants of signal intensity. However, in contrast to our findings, Doppler gain and power had limited influence. Mizushige et al.12 showed that the intensity of the power Doppler image was dependent on flow velocity and the wall motion filter under steady flow conditions within a single acrylic tube phantom. They used a saline solution containing 40% glycerol and cornstarch as the BMF, finding a high degree of correlation between the Doppler shift frequency, as measured by conventional pulsed wave Doppler, and the power Doppler image intensity at the center of the flow image. The power Doppler signal was also affected by the beam incident angle, which, along with velocity, was closely related to properties of the filter, suggesting this may be an important determinant at low velocity flow rates. While the wall motion filter did reduce the vascular indices as it was increased, the effect was less than that of many of the other variables, despite the use of the relatively low flow rate of 9.8 mL/s.
Most of these experiments examined 2D power Doppler data and very few have examined 3D power Doppler data, despite the fact that a flow phantom provides volumetric information. Li et al.13 reported a high degree of correlation between true flow rate and calculated flow rate using a combination of 3D ultrasound and 2D color Doppler in the assessment of flow convergence and quantification of regurgitant flow within an in-vitro flow system. Cloutier et al.14 used 3D power Doppler ultrasound to investigate the performance of different segmentation algorithms in the delineation of the lumen of a simulated, stenosed artery. They found that the wall motion filter, type of flow (steady or pulsatile) and flow rate all affected the accuracy of the power Doppler image in determining the degree of stenosis. Guo et al.15 used 3D power Doppler ultrasound to examine the degree of vessel stenosis in several wall-less vessel systems. The degree of stenosis was quantified with an overall accuracy of 8.3% and precision of 7.0% over a range of area reductions and different flow rates under both steady and pulsatile flow conditions. These studies used the 3D aspect more for spatial orientation than for volume calculation and none examined the 3D vascular indices calculated in our present study.
Quantification of the power Doppler signal within a 3D area using the histogram facility involves the detection and weighting of color voxels, otherwise known as 3D pixels, relative to the total volume being considered. VI reflects the relative proportion of color voxels within the user-defined area and FI their mean signal intensity. VFI represents a combination of these two measurements, obtained through their multiplication. These indices represent power Doppler information distributed within a volume during the period over which the acquisition was obtained. This introduces two significant variables: movement and time. Power Doppler is more sensitive to movement artifact than are other forms of Doppler imaging, but this may be limited by avoiding inappropriate transducer movement and by increasing the speed of data acquisition. When artifactual information is present it is usually readily distinguishable from true flow data by its non-physiological appearance. Time is more difficult to account for and, because all forms of 3D data acquisition involve the serial acquisition of 2D planes, the effect of the cardiac cycle must be considered. Data are being acquired during both systole and diastole, which produce different power Doppler maps; the power Doppler signal is increased during systole, due to vessel expansion and higher flow volume rates, and then reduces with diastole. In real time, the pulsatile nature of blood flow within the uterus is readily evident. However, during 3D data acquisition, the pulsatile effect is less apparent and appears ‘averaged’ throughout the sweep of the ultrasound beam. In this study, the fast sweep mode was actually associated with a significant fall in VFI relative to the medium and slow sweep speeds. It is possible, therefore, that the degree of any averaging of cardiac activity is reduced as the acquisition speed is increased. This is totally hypothetical, of course, and must be considered against the fact that only continuous flow was used in our study. Also, we used a Voluson 530 ultrasound machine, one of the first machines in this series. The newer Voluson systems (Voluson Pro, Expert and E8) offer improved power Doppler sensitivity and may further accentuate the effect of different Doppler settings, but should reduce the loss of information with faster acquisition speeds as more information is obtained within any acquired volume. Future work examining the effect of acquisition speed and variable pulsatile flow rates is warranted, particularly as this reflects the physiological environment in which 3D power Doppler is being applied clinically.
Another interesting observation was the generation of an FI value when no VI value was returned. This was evident during the gain experiment, when a FI of 10 was seen in association with a VI of 0 at a gain setting of 26.4 dB. We have consulted the GE Medical Systems Research and Development team in Kretz, who confirmed that this may occur when the number of color voxels is very low. Under these circumstances, the VI is displayed as 0 because the Doppler signal is below the precision of the display and cannot be seen. Power Doppler information is there, however, and an FI can be generated from this. There are other, more obvious, limitations to this study. We deliberately used a simple design for the flow phantom as this was intended as a preliminary study and we wanted to focus on machine settings only. We did not examine the effect of pulsatile flow or look at more complex arrangements such as multiple vessels with variable diameters. This will be the subject of future work which will also consider the effect of different ultrasound machines and flow phantom devices and designs. While our conclusions are therefore limited, and must be taken in the context of the study design, our results show that the 3D vascular indices are mostly affected in a predictable manner by serial adjustments in any Doppler parameter, emphasizing the importance of maintaining and reporting machine settings.
In conclusion, in this study we quantified the effect of different Doppler settings on the vascular indices obtained through the quantification of 3D power Doppler data and demonstrated that significant changes occur with subtle changes in these settings. We have shown how the vascular indices are affected by serial changes in different Doppler settings and that the indices follow a fairly uniform and predictable pattern with such changes. This work emphasizes the importance of maintaining Doppler settings between subjects to facilitate intersubject and intrasubject comparison within the research setting and clinical environment. Futher work is required to examine how pulsatile flow and more complex vessel arrangements affect these indices.