This study investigated the feasibility of a quantitative angiography method using time-resolved 2D DSA. The method was assessed for accuracy and precision in a phantom model and an in vivo porcine model. Our results indicate that qDSA allows the calculation of quantitative velocities, over a range of physiologic abdominal arterial velocities (Nakamura et al. 1989) in near real-time, that are both accurate and precise. Potential errors from angle projection and magnification were investigated and successfully corrected for, demonstrating the robustness of the technique. Quantitative velocities were computed in vivo using branches of the abdominal aorta and were found to be strongly correlated with an established quantitative MRI technique. The application of the proposed method was similar for both the experimental and in vivo studies, indicating the potential for clinical adaptation.
Prior intraprocedural quantitative imaging techniques have been described, including both 2D and volumetric techniques (Shaughnessy et al. 2018; Wu et al. 2018; Wang et al. 2010; Zhang et al. 2013; Hinrichs et al. 2016; Wang et al. 2016). Volumetric techniques including quantitative 4D transcatheter intra-arterial perfusion (TRIP) MRI and 4D DSA have the ability to provide flow, not just velocity, which can be important for interventions in which vessel diameters change. TRIP MRI lacks feasibility as it requires a complex hybrid angiography/MR suite only available at select institutions, and also requires significant time and cost. 4D DSA is useful for characterizing blood flow at baseline and upon completion of a procedure, but the long data acquisition times and susceptibility to motion artifacts make it less suitable for repeated use throughout a procedure.
Quantitative color-coded DSA is a commercially available 2D DSA-based technique (syngo iFlow; Siemens, Forchheim, Germany). However, it does not provide true arterial velocity or flow, rather it provides time-attenuation curves for specified points and color-coded vessel displays based on TOA or TTP. Such an analysis is prone to error and sensitive to changes in cardiac output, motion, total amount and duration of contrast medium administration, imaging parameters (eg, injection delays), and angiographic catheter position. The proposed qDSA technique takes advantage of spatial and temporal information along the vessel allowing for a more robust quantitative technique. The ability to characterize intraprocedural arterial velocity reductions during TAE using qDSA was recently demonstrated in a clinically relevant porcine liver model (Periyasamy et al. 2020). In that study, qDSA was compared to the commercially available iFlow. qDSA was able to quantitatively discriminate between embolization endpoints including sub-stasis, the desirable clinical endpoint previously correlated with improved overall survival (Jin et al. 2011). qDSA better characterized blood flow changes when compared with iFlow and qDSA endpoints correlated with tissue level changes. The results of that study, which included an ability to resolve a range of changes in arterial velocity, support the potential clinical role for qDSA not only for TAE, but also for other arterial interventions (e.g. angioplasty or stenting for peripheral arterial disease) where accurate assessment of changes in blood velocity are critical in determining the success of a treatment.
The qDSA technique described here could be easily translated to intraprocedural clinical workflows given that it would only require modification to image acquisition parameters. In the present study, all velocity calculations were performed within 2–3 min of data acquisition using the prototype MATLAB tool on a standard laptop (Intel Core i7-8550U 1.80 GHz CPU, 16 GB RAM). Further refinement of the technique and tool will likely lead to significant reductions in computational times, enabling near real-time determination of velocities repeatedly during procedures. Although qDSA is currently limited to blood velocity, further development may allow for calculated velocities to be converted to blood flow using forward projection techniques on pre-procedure 3D imaging (Hentschke et al. 2011). This would permit flow quantification in a manner more similar to 4D Flow MR or 4D DSA.
An accurate velocity calculation requires precise knowledge of both the distance the blood traveled and the time (or temporal shift). The measured distance will differ from the true distance if the projection is not orthogonal to the vessel segment. The proper projection angle can be determined from 3D imaging, either preprocedure or at the time of the procedure, and the true distance can be determined by placing an object of known dimensions into the vessel segment (eg a guidewire or catheter with radiopaque markings). While both projection angle and magnification corrections are important to achieve accurate absolute velocity values, relative velocity changes during a procedure (eg, pre- and post-intervention) can be calculated without incorporating corrections as long as the table position and projection angle are maintained. Furthermore, corrections for projection angle may be unnecessary for many abdominopelvic interventions given that many vessels are relatively straight, vessels can typically be laid out in AP or shallow oblique projections, and minimal variation (< 5%) in velocity was observed over a wide range of projection angles (± 15 degrees) in the phantom study. A minimum frame rate is required to achieve adequate temporal resolution for the higher velocity values encountered in clinical practice. In its current form, qDSA is associated with additional radiation dose (from additional high frame-rate scans). Radiation dose reduction strategies are currently being explored to facilitate incorporation of qDSA into standard clinical angiography workflows. In addition, preliminary studies with both preclinical and clinical datasets indicate that qDSA may also be viable with fluoroscopy, which would significantly decrease the exposure compared to subtraction angiography.
X-ray videodensitometric blood velocity methods have been previously described (Shpilfoygel et al. 2000). Many of these blood velocity techniques were developed for cerebrovascular interventions. Few have been developed or validated in the abdominal vasculature. Furthermore, an analysis tool for near real-time calculation has yet to be created for use in body interventions. We have developed a tool that allows the calculation of blood velocity from 2D DSA sequences within minutes with minimal user interaction. The graphical user interface allows common visualizations of DICOMS and TACs while providing quantitative blood velocity values from temporally and spatially segmented vessels of interest. The development of this qDSA velocity tool makes intraprocedural blood velocity calculations feasible and more readily translatable to clinical workstations.
Our study had several limitations. Our preliminary in vitro testing was performed in a bifurcation phantom with three segments, all relatively linear and of constant diameter. The velocity in this phantom is primarily laminar, but the larger outlet does contain a region of recirculation. The abdominal arteries have more tortuosity and variation in vessel diameter which can lead to increased turbulent velocities and disruption in pulsatile signal. Additionally, we used a pump that generated a repeatable pulsatile signal, which may not entirely represent the hemodynamic heterogeneity of abdominal vasculature. Despite the limitations of the phantom model, the results of the in vivo experiments suggest the technique is robust and accurate with more complicated vessel geometries. In our in vivo testing, the MR velocity data was acquired without an intra-arterial injection requiring an additive velocity correction before a direct comparison to qDSA velocity could be made. To replicate the DSA injection during MRI, catheter placement could be completed under fluoro guidance and then the animal could be moved to the MRI scanner. However, for a multi-vessel analysis, this would require a minimum of 5 trips between imaging modalities with the risk of catheter movement occurring during each transport. Additionally, the MR scans would need to be adjusted to capture near real-time velocities, limiting the scans to 2D techniques which would no longer allow averaging along the vessels of interest.