Year: 2004
Type: article
Abstract: Complete fetal assessment should ideally encompass both structural and functional components. Accurate assessment of regional blood flow, in particular impaired or redistributed flow, is necessary for functional fetal assessment. Various methods of evaluating perfusion in the adult have been well summarized by Jansson et al.1, but the mainstay in the fetus has been Doppler ultrasound2. Doppler ultrasound allows the evaluation of moving particles within the fetal circulation but has not yet been proved useful for direct measurement of regional perfusion. It would appear simple to use quantified Doppler to evaluate flow volumes in single large vessels, by combining accurate gray-scale measurement of vessel diameter with spectral Doppler calculation of the speed of movement of the blood cells within that vessel3-5. However, even for isolated large vessels, inaccuracies in both the gray-scale and the Doppler component have so far precluded accurate volume flow measurements, with the pulsatile expansion of vessels such as the fetal aorta being in the range of 10–15% of the minimum diameter3. This becomes significantly more complex when extrapolated to tissue perfusion, where blood vessels are tortuous and multidirectional, prohibiting accurate vascular diameter measurement or angle of insonation correction for the Doppler frequency shift. For this reason fetal vascular assessment has so far relied upon the measurement of indices of resistance to blood flow in isolated blood vessels, including the systolic/diastolic ratio, the resistance index and the pulsatility index. The advantage of these measurements has been that an alteration in the angle of insonation will result in an equal increment in both the numerator and the denominator, and thus the ratio remains consistent. Based upon these ratios, comments on tissue perfusion and vascular compliance may be made, as an extrapolation from the perceived resistance to flow; the most studied cases are the progressive changes in the indices of the umbilical and middle cerebral arteries and ductus venosus in fetal growth restriction6. A newer option in the search for a tool to evaluate regional perfusion is power Doppler. All machines that produce color Doppler through autocorrelation inherently have power Doppler capability, as both power and spectral modes are derived from the same information7. Whereas spectral (velocity-based) Doppler examines and displays frequency shifts, power Doppler instead displays details of the intensity or amplitude of these frequency shifts above a certain threshold value (the noise filter setting). While the frequency of the Doppler shift from a given particle (such as a red blood cell) is related to the cosine of the angle of insonation, the amplitude of the scatter is not. This means that power Doppler imaging is (essentially) angle-independent8. Background noise is more easily filtered out, as it is always of low amplitude9, allowing the display dynamic range to be significantly increased10, with higher gain levels9. These combined attributes make power Doppler more suited to evaluation of multidirectional flow and hence perfusion. It is for this reason that various research groups are investigating power Doppler as a tool for quantification rather than observation. The development of these techniques have been well documented11-18. Power Doppler is ideally suited to measurement of values that approximate to perfusion, though are not, by definition, perfusion measurements themselves19. Instead it gives a measurement of the number of scatterers (predominantly red blood cells) and their amplitude within tissue. In-vitro experiments have shown power Doppler intensity to correlate with fluid flow8, 20, 21. Regional vascularity as measured with power Doppler may be assessed by counting the number of color pixels and the intensity of color within these pixels in a given region on power flow mapping. This may be performed by software within the ultrasound machine itself, or clips or images may be exported from the ultrasound machine for offline digital analysis. The most commonly used techniques are to compare the density of colored pixels (percent of total area), the number of visible vessels (per unit area) in a region of interest, or the summed pixel values and compare with normal tissue1. One of the concerns with quantification of power Doppler images is that while numerical information can be ascribed to a given image or clip in this fashion, outside of the examination in question measurements of power Doppler may have little meaning, as they are based on purely arbitrary indices. Any factor that alters the color pixel count could affect vascularity assessment, including effects of attenuation (maternal habitus, fetal depth, intervening tissue type) and machine settings (aperture size, gain, scale, filter and persistence). Though there is time-gain-compensation in power mode Doppler, it is unlikely to be effective over the whole field of view. In an in-vitro model the flow area derived from power Doppler was affected by changes in pulse repetition frequency, color filter and frame rate22. The need for care with alterations in machine settings was highlighted in another study which examined the effects of filter, scale and angle of insonation for a single large vessel in an experimental setting11. In addition, while it may be possible to standardize for scale, gain and depth, these factors are not necessarily linear. Calculated indices that ignore the effects of non-linear signal processing cannot be expected to be linearly related to indices of the circulation22. While many modern ultrasound machines, including those designed for three-dimensional and 4D imaging, allow the extraction of statistics and information related to regional amplitude scattering, studies using these need to be interpreted with caution. The reason for this is that, in the absence of a local standardizing value, all measurements are related to an arbitrary scale. This means that they cannot be extracted and compared between examinations on the same or other patients. Because 3D and 4D machines can automatically measure indices of regional amplitude, there is a large developing body of literature on their use to describe perfusion. The only situation in which these indices may be extracted and used between cases is in transvaginal gynecological sonography, when the transducer may be directly adjacent to the area under examination, thus minimizing attenuation. Only from papers in which rigorous attention has been paid to consistency of machine settings and distances can any meaningful information be gained23, 24. Papers in which quantification has taken place without standardization of values should be disregarded25. A large body of work has been published by a Swedish research group, looking at quantification of power Doppler using non-standardized techniques26-31, showing an increase in power Doppler signals from the placenta, fetal lungs, spleen, liver and kidney up to 34 weeks' gestation, followed by a decrease in all but the spleen, which remains constant. Fetal cerebral power Doppler intensity and brain/lung ratio also increase towards the end of pregnancy with an increased cerebral power Doppler signal and brain/lung ratio in high-risk pregnancies. In a large study of 180 high-risk pregnancies, semi-quantitative computer analysis of power Doppler using pre-fixed Doppler settings was evaluated in the placenta, brain, lung, liver, kidney and spleen of the fetus31. There was a claimed relationship between fetal brain-tissue blood flow and perinatal outcome, though this was measured in less than half of the cases and was compared with published non-standardized reference ranges from the same group29. Whilst pre-fixed Doppler settings were used, these would have been influenced by attenuation and the authors state that depth of scanning differed between examinations, which would have limited their conclusions. A simple technique to overcome the limitations of attenuation and machine settings has been proposed in two stages by a team based at the University of Michigan. In the first stage of their proposal they described the use of a region of known 100% vascularity, such as the center of a large blood vessel, to define the maximum permissible value for vascular amplitude scattering21. Above a given threshold, power Doppler values for 100% blood amplitude should be constant and independent of velocity, allowing a reproducible potential maximum value to be calculated. Any regional measurements could then be compared with this local maximum as a 'fractional' perfusion of this tissue, expressed on a scale from 0 to 1 or as a percentage value against the maximum of 100%. If the standardization point was measured at the same depth and under the same imaging conditions, then any alteration in machine settings should alter the signal from both the region-of-interest and the maximum reference point, resulting in a constant fraction. Exceptions to this would be in regions of extremely low amplitude flow, where the Doppler filter settings might remove some of the lowest readings, preferentially affecting the numerator and not the denominator in the fraction. When these values are expressed in relation to tissue volume they may be described as 'fractional moving blood volumes' or FMBV. The components of the term FMBV are that blood volume (amplitude of individual scatterers of ultrasound) is measured instead of velocity, that only the moving component of blood (as detectable by Doppler frequency shifts with sufficient amplitude) are measured, and that the measurement is shown as a regional fraction standardized to the maximum potential detectable amplitude with 100% blood. The second important paper from the team in Michigan in the development of quantified power Doppler examined the influence of clumped red blood cells, or rouleaux, on the value ascribed to 100% flow (the standardization value)32. It might be predicted that the central regions of adult blood vessels would have a constant power Doppler measurement, corresponding to a 100% concentration of blood, giving a flattened amplitude profile in cross-section. In fact, the vascular profile across large vessels was found to continue to a central peak. The reason for this was the presence of rouleaux, which were very high-amplitude ultrasound scatterers, giving an artificially high maximum value across the vascular profile. In order to compensate for this finding, they devised mathematical techniques for separating out the two populations of scatterers: red blood cells and rouleaux. Their technique involved looking at the histogram of Doppler amplitude across the large blood vessel, the cumulative power distribution function (CPDF). Using this graphical plot, they showed with the aid of an ultrasound phantom and in the renal cortices of healthy volunteers that two techniques could be used to estimate the true (intermediate) value associated with the amplitude scattering of 100% red blood cell concentration. Values above this in large vessels could be normalized to 1 and values below this could then be adjusted accordingly to be a fraction of 1. These techniques involved finding a point of inflexion in the cumulative histogram of pixel values across the vessel, between the power Doppler amplitudes of red blood cells and rouleaux. This value could then be used to approximate to a true maximum for 100% red blood cells only. They used two separate methods, the first drawing tangent lines to the CPDF and identifying their point of intersection, and the second rotating the CPDF to find a point of inflexion. Thus, in the adult, they were able to overcome the effects of rouleaux formation and determine the standardization value accurately. The potential for two- and even three-dimensional measurement of regional perfusion is enormous. However, as with all such Doppler research, it is essential that these new techniques are proven to be reproducible and that the techniques are shown to correlate accurately with the perfusion that they are intended to represent. In this issue of the Journal, two papers from a team based at the University of Lund, Sweden provide useful background for future research into the quantification of power Doppler. A previous publication from this group has outlined methodological aspects of evaluation of fetal lung perfusion using FMBV33. In this work they evaluated 29 uncomplicated singleton pregnancies, with a coefficient of variation for mean FMBV from six stored ultrasound frames from each patient of 0.14. They also demonstrated that neither FMBV nor mean pixel intensity had any depth dependence, though the calculated standardization factors had a statistically significant depth dependence (P < 0.01). Interestingly, there was no measured difference between FMBV in systole and in diastole, with the assumption that a relatively high persistence setting compensated for fluctuations in true perfusion. In the first of the papers in this issue, Hernandez-Andrade et al. examine the relationship between measured FMBV and the 'gold standard' of perfusion evaluation, the use of radiomicrospheres (RMS) in the adrenal glands of exteriorized lamb fetuses34. Examining the adrenal glands of nine such fetal lambs, they induced asphyxia with redistribution by total umbilical cord occlusion and compared this group with an observational control group. Three RMS injections with different labeled isotopes were given before occlusion, 5 min after cord occlusion and when the mean arterial blood pressure decreased below 25 mmHg. Power Doppler evaluation of adrenal perfusion was performed in a water bath, ten consecutive frames being selected from each sequence. After sacrifice, nuclear activity was calculated for each isotope and the adrenal blood flow was calculated by comparing the number of RMS in reference samples with the number in the adrenal gland, allowing adrenal flow to be calculated in mL/min/kg. The correlation between FMBV and RMS was good (median, r = 0.90; range, 0.43–0.99), which is extremely valuable to researchers in this field, showing that although FMBV does not directly measure perfusion, it can accurately approximate perfusion. More importantly, it helps to show that alterations in perfusion are reflected by alterations in FMBV. In the second paper from this group, Hernandez-Andrade et al. evaluate the reproducibility of FMBV evaluation by two experienced observers in 20 normally grown singleton fetuses at 32–35 weeks' gestation35. While FMBV could only be evaluated in 85% of the cases by both observers, the intraclass correlation coefficient for Observer A was 0.92 (95% CI, 0.78–0.96) and for Observer B it was 0.90 (95% CI, 0.74–0.96). The interclass correlation coefficient for agreement between the two observers was 0.92 (95% CI, 0.84–0.95). These results indicate a high degree of reproducibility for this technique when it is in the correct hands. The CPDF and FMBV techniques have been shown to work well in the hands of the team of Hernandez-Andrade et al. The latter technique or technical description is necessary to overcome the effects of attenuation and machine settings. The former technique was proposed as a variation in the adult to overcome the artificial increase in value for 100% created by the presence of rouleaux. If rouleaux are present to any significant degree in the human fetus then the CPDF algorithm is necessary to correctly normalize values ascribed to perfusion. However, if the fetus does not form rouleaux and has different rheological properties from the adult, then this may not be necessary, as suggested by some neonatal literature36. Clearly, further work may be necessary to resolve this issue. These two papers from Hernandez-Andrade et al. lay significant foundations for future research into quantification of power Dopper by showing the correlation between the new index of FMBV and true perfusion as measured by RMS, and the reproducibility of its measurement. The logical extension of these techniques is into 3D standardized techniques for FMBV measurement. This would remove one potential source of error, the difficulty in identifying a reproducible anatomical plane for imaging, though it would introduce a further potential error as the transducer sweeps across a region of interest while the cardiac cycle is fluctuating. Hopefully, progress in this area will focus on papers examining the reproducibility and validity of these techniques before reference ranges for normal and abnormal 'perfusion' in the human fetus are determined.
DOI: 10.1002/uog.1037
Author: A. Welsh
Institution: Royal Prince Alfred Hospital
Language: English
Subfield: Obstetrics and Gynecology
Field: Medicine
Domain: Health Sciences
Open Access status: bronze
OpenAlex ID: W2003359829