Comprehension of power Doppler quantification and a missed opportunity for landmark research

Reading the November Opinion piece regarding power Doppler reminded me that the concepts underlying this area of our discipline remain extremely hard to fully comprehend1. I have many concerns regarding Dr Martins' use of this technology, the majority of which have been addressed in previous correspondence2, 3. There were several inaccuracies within the Opinion regarding fractional moving blood volume (FMBV)4. The statement that ‘only the information related to Doppler signal intensity is normalized’ is incorrect as FMBV generally relates to an integrated intensity over an area of tissue, thus including a component relating to the proportion of vascular voxels or pixels4. Additionally, FMBV is intended to standardize for attenuation or machine settings. In comparison, technology such as VOCAL™ (Virtual Organ Computer-aided AnaLysis), in which the gain setting can affect measured values that are taken to represent ‘perfusion’, has been accepted without question for too long2, 3. As we have stated previously, the manufacturer itself does not support the use of VOCAL to approximate flow measurement3. The fact that there have been ‘> 100 papers published with VOCAL’1 does not attest to the scientific merit of this technology, but instead merely reveals the ‘push-the-button’ ease of a system in which no information is made available about its algorithmic derivation. The end-point of research should not be publication volume, it should be enhanced care, and the intermediate points should include meaning and validity. It is perfectly reasonable in an Opinion piece to have and express individual views, and it is part of the ‘cut-and-thrust’ of academia to critique the views of others; however, one must take care to not inadvertently attribute falsely methodology that may discredit the work of others1. Martins describes a paper in which we found the vascularization index, flow index and vascularization flow index to be hugely variable tools for evaluation of flow within the placenta. This was unsurprising to many who have attempted to use this technology, though potentially disturbing to others basing academic careers upon this technology5. When asked to explain some reasons why one might find such variability, we responded by describing how individual vessels may influence the measurement of VOCAL indices even within a stable 3D dataset3. In order to make this as simple to understand as possible, we suggested that researchers might store datasets using standard methods. Having done so, they might deliberately choose volumes from within these datasets, some containing large vessels and others excluding large vessels, in order to demonstrate the influence of including large vessels in the region analyzed on the perfusion value obtained. This was an exaggerated description to highlight the potential flaws in selection of regions for analysis. It was not, as Martins seems to suggest, the technique that we used in our publication on placental VOCAL to show reliability of data acquisition. We were clearly attempting in our letter to explain as simply as possible how one might create inaccuracy even from within a static 3D dataset. From our correspondence and previous research it is quite obvious that we would not have committed this basic research error4, 6. From the same issue, Morel et al.7 should be commended on navigating the ethical, practical and collaborative aspects of such a technically complex animal model to evaluate the relationship between uterine artery flow and placental perfusion as measured by VOCAL. They successfully anesthetized pregnant sheep, exteriorized the pregnant horn, dissected out the common uterine artery and ligated all other uterine arterial supplies, then used a quantitative sensor and a controllable vascular occlusion system. One would anticipate that the ‘simple’ part of the study would then be to take a large number of readings from multiple placentomes, and to perform thousands of analyses to prove convincingly the point of accuracy. Surely the authors would do this if they truly believed that VOCAL was accurate? Instead the paucity of data resulting from this complex model may have wasted this valuable animal research opportunity, with fewer than seven datasets analyzed per sheep (69% of those stored). The authors attribute their small amount of data to difficulties in maintaining steady uterine perfusion, though they describe concurrent measurement of flow with each stored dataset, which would have allowed direct statistical comparison. Their methodological findings contrast with those of other authors who have not had the same trouble maintaining steady flow in the ovine uterine artery8, 9. The dearth of data extends from acquisition to analysis, with minimal measurements (only two) being taken at each rotation angle, seemingly to avoid variability. The analysis at 30° appears to be deliberately reducing the accuracy of tracing the contour of the maternal side of each placentome by reducing the number of tracings per volume, resulting in an unnecessarily coarse analysis. Morel et al. discuss the ∼80 placentomes per ovine placenta but have overlooked the opportunity to compare VOCAL (as a marker of regional variation) between these placentomes. Thus the applicability of their measurements to actual perfusion, or even the true relation to uterine arterial flow, is weakened. It is hard to imagine why the imaging analysis was so poorly conducted given the difficulty, expense and time that such an animal model must have entailed. There is clearly much further work that needs to be performed to investigate the meaning and applicability of quantified power Doppler, in both two and three dimensions. It is particularly unfortunate that this is required after widespread adoption of the commercial simplistic VOCAL technology with its ‘> 100 papers published’1. A.