Inapplicability of fractional moving blood volume technique to standardize Virtual Organ Computer-aided AnaLysis indices for quantified three-dimensional power Doppler

Authors

  • A. W. Welsh,

    Corresponding author
    1. Department of Maternal-Fetal Medicine, Royal Hospital for Women, Randwick, NSW, Australia
    • School of Women's and Children's Health, University of New South Wales, Randwick, NSW, Australia
    Search for more papers by this author
  • S. L. Collins,

    1. The Nuffield Department of Obstetrics and Gynaecology, University of Oxford, Oxford, UK
    2. The Fetal Medicine Unit, John Radcliffe Hospital, Oxford, UK
    Search for more papers by this author
  • G. N. Stevenson,

    1. Institute of Biomedical Engineering, Department of Engineering Science, Oxford University, Oxford, UK
    Search for more papers by this author
  • J. A. Noble,

    1. Institute of Biomedical Engineering, Department of Engineering Science, Oxford University, Oxford, UK
    Search for more papers by this author
  • L. Impey

    1. The Nuffield Department of Obstetrics and Gynaecology, University of Oxford, Oxford, UK
    2. The Fetal Medicine Unit, John Radcliffe Hospital, Oxford, UK
    Search for more papers by this author

Correspondence to: Prof. A. W. Welsh, Department of Maternal-Fetal Medicine, Royal Hospital for Women, Locked Bag 2000, Barker Street, Randwick, NSW 2031, Australia (e-mail: alec.welsh@unsw.edu.au)

Abstract

Objective

To determine whether the technique of fractional moving blood volume (FMBV) is applicable to Virtual Organ Computer-aided AnaLysis II (VOCAL II™)-based indices to quantify three-dimensional power Doppler ultrasound (3D-PDU) by investigating the effect of gain level on the indices measured at a possible reference point for standardization.

Methods

Ten women with singleton pregnancy between 33 + 3 and 37 + 5  weeks' gestation were recruited. The optimal position for 3D acquisition of cord insertion into the placenta was identified and static 3D-PDU volumes were acquired using consistent machine configurations. Without moving the probe or the participant changing position, successive 3D volumes were stored at −3, –5, –7 and −9 dB and at the individualized sub-noise gain (SNG) level. Volumes were excluded if flash artifact was present, in which case all five volumes were reacquired. Using 4D View software, the cord insertion was magnified and the smallest sphere possible was used to measure vascularization index (VI), flow index (FI) and vascularization flow index (VFI). The associations between VOCAL indices and gain level were assessed using Pearson's correlation coefficient.

Results

VOCAL indices for cord insertion correlated poorly with gain level, whether fundamental or relative to SNG level (R2 = 0.07 and 0.04, respectively). VI was consistently 100% and mean FI and VFI were 99.5 (SD, 0.57), with all values > 97 irrespective of gain level.

Conclusions

Whilst previous work has shown that gain correlates well with placental tissue VOCAL indices, the correlation between gain level and VOCAL indices in an area of 100% vascularity at the cord insertion is poor. Regions of 100% vascularity appear to be artificially assigned a value approaching 100% for all VOCAL indices irrespective of gain level. This precludes using the technique of VOCAL indices from large vessels to standardize power Doppler measurements and the FMBV index is therefore not applicable to image analysis using VOCAL.

Introduction

Power Doppler ultrasound (PDU) signals represent the concentration of moving ultrasound scatterers (red blood cells) and the amplitude of their scattering. The potential for PDU to be quantified to produce indices relating to vascularity or perfusion has been discussed for a long time, generating a wealth of research papers concerning both two-dimensional (2D) and three-dimensional (3D) scanning[1-3]. Quantification of power Doppler relates to the generation of an integral (or discrete sum) of the amplitudes of ultrasound scattering within an area of interest, also described as the ‘area under the curve’. The resulting integral potentially may be related to the vascularity of that tissue. However, machine settings[4, 5], maternal habitus and attenuative effects along the beam path may all alter the PDU signal[6-8]. In order to provide quantifiable measurement of ‘vascularity’, either between subjects or longitudinally, standardization of any PDU measurement must be performed to compensate for these effects. Without this intervention, any recorded PDU measurements can be disregarded because of error introduced by these factors.

In two-dimensional imaging, standardization has been achieved by using a technique known as fractional moving blood volume (FMBV)[9], the application of which in fetuses has been outlined in a previous opinion piece[10]. In brief, this technique involves selection of a region within the ultrasound image that is known to have 100% vascularity (e.g. a large blood vessel such as the fetal aorta) which is used as the ‘standardization point’ against which other readings are compared. It makes the assumption that alterations in the machine and beam path factors will alter signals from the region of interest and the standardization point to a proportionally equal degree. A hypothetical example showing consistent calculation of the FMBV index at different gain levels is shown in Figure 1.

Figure 1.

Hypothetical example showing the influence of an increase in gain setting on the measured value of integrated power Doppler amplitude for tissue (··×··) and for a proposed standardization point (··▪··). In this case, as both the numerator (tissue value) and denominator (value at standardization point) are increased by gain, the ratio remains 1 : 4 and so the fractional measurement is 25%, irrespective of gain setting.

It has been suggested recently that a standardization technique based upon the Virtual Organ Computer-aided AnaLysis II (VOCAL II™) method, a part of the 4D View™ (GE Healthcare Ultrasound, Milwaukee, WI, USA) volume visualization application, could provide standardized PDU measurements based upon the standard VOCAL™ indices: vascularization index (VI), flow index (FI) and vascularization flow index (VFI)[11]. We have demonstrated recently a clear linear relationship between PDU gain level and VI/VFI values in vascularized areas of placental tissue[12] and thus hypothesized that a change in PDU gain would alter the proposed standardization value accordingly. Therefore, keeping all other settings constant, we sought to evaluate the effect of PDU gain alteration upon the measured value that should act as the standardization point.

Methods

The data for this study were collected as part of another, larger study with UK National Health Service Research Ethics Committee (REC) ethical approval (REC ref.: 08/H0604/163). Women over the age of 16, with a singleton pregnancy and body mass index (BMI) <30  kg/m[2] were recruited prospectively for a study involving additional antenatal ultrasound examination. Gestational age was calculated from crown–rump length measurement at the first scan, performed between 11 + 0 and 13 + 6  weeks' gestation[13].

The examinations were undertaken by a single operator (S.C.) using a Voluson E8™ (GE Healthcare Ultrasound) ultrasound machine with a RAB4-8-D 3D/4D curved array transabdominal transducer (4–8.5  MHz). All scans were performed between 33 and 37  weeks' gestation. After confirmation of viability and identification of placental position, the optimal position for 3D acquisition of cord insertion into the placenta was identified. A static 3D power Doppler volume of the cord insertion was acquired using a previously saved machine configuration including wall motion filter low 1 and pulse repetition frequency 0.9  KHz (thus ensuring that identical machine settings were used for each subject) and the power Doppler gain was set at −3  dB. Without moving either the participant or the probe, the power Doppler gain was quickly changed to −5  dB and a second static 3D image was captured. This technique was repeated for gain settings of −7 and −9  dB and the individualized ‘sub-noise’ gain (SNG) setting for that participant as recently described[12]. Apart from the power Doppler gain, all machine settings were kept constant for all scans. Throughout data acquisition the participant was asked to remain as still as possible. To ensure that all five captured volumes contained exactly the same anatomy, if either the probe or the woman moved the sequence of five scans was repeated. The volumes were then checked for ‘flash artifact’ (secondary to fetal or maternal movements) and, if present, all five scans were rejected and the sequence of scans was repeated. The data were then stored using Sonoview and analyzed offline on a personal computer using VOCAL II™ on 4D View™.

Each static volume was opened in 4D View and analyzed by magnifying the cord insertion as much as possible and placing the smallest possible sphere in the center of the vessel (< 1  cm3). This is demonstrated in Figure 2. The sample volume was checked to ensure that it was the same size for each of the five volumes. This meant that, for each participant, the only difference in the five samples analyzed was the setting of the power Doppler gain. The VOCAL histogram was then viewed and values for VI, FI and VFI were recorded. Statistical analysis was performed using PASW Statistics v18 (SPSS Inc., Chicago, IL, USA), including evaluation of correlation between gain setting and VOCAL indices using Pearson's correlation coefficient.

Figure 2.

Analysis of a three-dimensional power Doppler ultrasound volume, showing selection of a central portion of the cord insertion, representing 100% vascularity, and the resulting Virtual Organ Computer-aided AnaLysis (VOCAL) indices. The sphere represents the region of interest for the standardization value (closed arrow), while the histogram box (open arrow) shows the values for vascularization index (VI), flow index (FI) and vascularization flow index (VFI).

Results

Ten women with singleton pregnancy were recruited between 33 + 3 weeks and 37 + 5  weeks' gestation. The cord insertion was clearly visible in all participants. Their body mass indices (BMIs) ranged between 22.0 and 27.2 (mean 25.1) kg/m2. The VI (reflecting the number of color-containing voxels) was 100% for all 50 volumes, indicating true pure color/flow sampling. The FI and VFI were therefore equal and had a mean value of 99.5 (SD, 0.57; all values > 97), which we call (V)FI.

Figure 3a shows a plot of fixed gain settings against (V)FI for each case with a truncated scale bar. The correlation between fixed gain setting and (V)FI is shown in Figure 3b (R2 = 0.07), and that between individual gains relative to the SNG level and respective (V)FI is shown in Figure 3c (R2 = 0.04). In our recent study, a comparable decrease in gain resulted in a mean alteration in VFI of 57.5% (range, 12.7–76.8%) in placental tissue[12].

Figure 3.

Influence of gain setting (dB) on vascularization flow index/flow index ((V)FI) values at fixed gain settings: (a) plot of gain setting against (V)FI for each subject; (b) correlation between fixed gain setting and (V)FI (R2 = 0.07); and (c) correlation between gain relative to sub-noise gain value and (V)FI (R2 = 0.04).

Discussion

The influences of machine settings and beam path mean that PDU readings ascribed to degrees of vascularity need to be standardized[14]; otherwise, even use of the gain button may change the proposed vascularity of tissues[12]. The best way to standardize assessment is to take a measurement of PDU within the tissue under examination that is relative to the value for 100% vascular amplitude within the same frame or volume (the standardization value). In order for this ‘standardization point’ or value to have meaning, it must be influenced to the same degree as its respective tissue, i.e. a 30% increase in measured power Doppler within tissue is matched by an equivalent 30% increase in the value for the standardization point (as shown in Figure 1); the fraction between the two would therefore remain constant.

Suggested approaches to standardize PDU in this way include the FMBV technique[9]. The bulk of fetal 2D FMBV work has been carried out by Hernandez-Andrade et al., using the specific mathematical algorithm described by Rubin et al.[15] and implemented in MATLAB® (Mathworks, Natick, MI, USA)[16, 17]. This method calculates the value for true 100% vascularity in an image using the cumulative power distribution function (CPDF)[15]. The CPDF corrects for the influence of rouleaux formation, as seen in adult blood, which would otherwise create an artificial value for ‘100% maximum’, higher than that seen for true 100% red blood cell concentration[15]. Other work has suggested that rouleaux are not present in fetal blood, making it simpler to measure directly from the center of an adjacent large vessel, e.g. the aorta or vena cava, the value to ascribe to 100% vascular volume or 100% fetal red blood cells[18].

The optimal technique remains to be clarified. Nevertheless, 2D standardization is easily achievable and it is tempting to try to extend this technique to 3D imaging as, without standardization, the validity of 3D vascular indices is being questioned increasingly[3, 8].

In 3D ultrasound imaging, semi-quantitative indices may be generated, including those produced by VOCAL: the VI showing how many voxels (3D pixels) contain color, the FI showing a weighted average value of color in the color-containing voxels and the VFI showing the product of the VI and FI[19]. Of these indices, the VFI most closely approximates the moving blood volume, though it is not actually the true integral of the power spectrum as originally described, but is instead related to a weighted mean value. As generated by VOCAL, the VI, FI and VFI are unstandardized. We have demonstrated recently a linear relationship between the PDU gain level (actually the ‘volume’ control) and these measured indices in tissue[12], underscoring the need for standardization. In our previous study, alteration in gain altered placental tissue VOCAL indices by up to 57% within the gain range employed by others[20]. If a standardization point were to be determinable using VOCAL, we would need to see the same relative change in indices at this point as that seen within the tissue.

We set out in this study to demonstrate an equivalent change in standardization value with alteration in gain settings, in order to confirm that VOCAL could be ‘internally standardized’ to create a 3D equivalent of the FMBV. We ensured that the same anatomical volume was captured five times consecutively, in the same women and without moving the ultrasound probe, with the only change being in the power Doppler gain setting. We anticipated that an area of 100% vascular volume would generate values for the VOCAL indices that changed with gain in keeping with alterations seen in tissue[12]. However, somewhat surprisingly, our findings are that, at the point of 100% vascularity, there is little correlation between VOCAL indices and gain levels, whether expressed as fundamental values or relative to the SNG level. All readings approximate 100% and show only a very slight degree of change with gain compared to those presented in our recent paper on placental indices[12]. The mean increase in value for (V)FI, from a gain level of −9 to −3 dB at the ‘standardization point’ in our current study was 0.3%. This compares with an alteration in placental tissue of 57.5% (range, 12.7–76.8%). Whilst the participants in the previous study are not the same as those in the current study, we feel that differences in subjects could not account for such a difference in results. In our study we have deliberately chosen gain levels equivalent to those used in contemporary VOCAL studies of the placenta[20].

It appears that the FMBV technique of standardization to a local value for 100% vascularity is inappropriate using 3D-PDU unless further refinements are developed, due to the consistent 100% levels observed regardless of changes in gain. These findings raise the question of what underlying signal processing is taking place within the ultrasound machine between echo return and display. From examination of histogram values it appears that the machine may be artificially truncating voxel values, losing values or stretching the maximum to 100%. The precise meaning of all VOCAL indices must therefore come under question if the VOCAL scale is being altered in this way. The VOCAL system was introduced primarily for visualization and manipulation of 3D images and volumes, and thus the measurements VI, FI and VFI may never have been intended to be used to quantify perfusion or vascularity. The use of VOCAL for quantification of power Doppler does not appear to be officially endorsed by the manufacturers, despite the wealth of published literature in this area. It is essential that errors present in VOCAL as a tool for estimation of vascularity or perfusion are not compounded by an inappropriate extrapolation to a false 3D equivalent of FMBV.

As gain alters VOCAL values within tissue[12, 21] but does not alter the value for regions of 100% vascularity that would be used for standardization, 3D FMBV using VOCAL is a technical impossibility. The term FMBV (or similar terminology) should therefore never be applied to any measurements based upon VOCAL as this may be misleading and cause confusion with existing, appropriately standardized 2D FMBV work.

Acknowledgments

S.L. Collins and G.N. Stevenson were supported by the Oxford Partnership Comprehensive Biomedical Research Centre with funding from the UK Department of Health NIHR Biomedical Research Centres funding scheme.

Ancillary