Background The hemodynamic evaluation of cerebral arteriovenous malformations (AVMs) using DSA has not been validated against true flow measurements.
Objective To validate AVM hemodynamics assessed by DSA using quantitative magnetic resonance angiography (QMRA).
Materials and methods Patients seen at our institution between 2007 and 2016 with a supratentorial AVM and DSA and QMRA obtained before any treatment were retrospectively reviewed. DSA assessment of AVM flow comprised AVM arterial-to-venous time (A-Vt) and iFlow transit time. A-Vt was defined as the difference between peak contrast intensity in the cavernous internal carotid artery and peak contrast intensity in the draining vein. iFlow transit times were determined using syngo iFlow software. A-Vt and iFlow transit times were correlated with total AVM flow measured using QMRA and AVM angioarchitectural and clinical features.
Results 33 patients (mean age 33 years) were included. Nine patients presented with hemorrhage. Mean AVM volume was 9.8 mL (range 0.3–57.7 mL). Both A-Vt (r=−0.47, p=0.01) and iFlow (r=−0.44, p=0.01) correlated significantly with total AVM flow. iFlow transit time was significantly shorter in patients who presented with seizure but A-Vt and iFlow did not vary with other AVM angioarchitectural features such as venous stenosis or hemorrhagic presentation.
Conclusions A-Vt and iFlow transit times on DSA correlate with cerebral AVM flow measured using QMRA. Thus, these parameters may be used to indirectly estimate AVM flow before and after embolization during angiography in real time.
- Arteriovenous Malformation
- Blood Flow
- Magnetic Resonance Angiography
Statistics from Altmetric.com
Cerebral arteriovenous malformations (AVMs) are vascular abnormalities consisting of direct connections from arteries to veins through an intervening network of low-resistance vessels called the nidus, rather than through normal capillary beds, thereby resulting in altered hemodynamics.1 Consequently, the characterization of cerebral AVM hemodynamics is of significant interest in the evaluation and management of AVMs. Indeed, the management of unruptured AVMs is debatable, which is largely owing to our lack of understanding of the anatomical and hemodynamic features of AVMs that are associated with increased hemorrhage risk.2–4 DSA provides anatomical imaging of AVMs and also a real-time assessment of hemodynamics due to rapid acquisition of quantitative data in the neuroendovascular surgery suite. Thus the information obtained from DSA may significantly affect the treatment of patients with AVMs.5–8
The hemodynamic evaluation of AVMs using DSA has been reported in the literature5 ,6 and can be performed in conjunction with commercially available software, such as syngo iFlow (Siemens Healthineers, Erlangen, Germany) and AngioViz (GE Healthcare, Chicago, USA).9 ,10 However, these techniques have not been validated against true flow measurements. Here, we introduce a new method of DSA image processing to determine cerebral AVM hemodynamics and correlate arterial-to-venous time (A-Vt) obtained with this technique and iFlow transit times with AVM flow measured by quantitative magnetic resonance angiography (QMRA).
Materials and methods
Following institutional review board approval, records of patients with a cerebral AVM evaluated at our institution between 2007 and 2016 were retrospectively reviewed. Patients were included if they had DSA and QMRA studies obtained before any treatment. Patients with infratentorial AVMs were not included in this study.
Nidus volume was calculated based on measurements obtained from MRI using the ABC/2 method. Presence of angioarchitectural features was determined from DSA, including arterial ectasia, intranidal fistula, intranidal aneurysm, venous ectasia, venous stenosis, and venous varix. Ectasia and stenosis were defined as at least a 50% increase or reduction in the original vessel diameter observed in any portion of the vessel.11
All angiograms were obtained using the same protocol for each patient. Specifically, a total of 12 mL of the iodine-based contrast agent iohexol (300 mg/mL, Omnipaque, GE Healthcare, Chicago, Illinois, USA) was injected into the cervical internal carotid artery (ICA) through a transfemoral approach. Contrast was injected by a power-contrast injector (Medrad, Bayer HealthCare, Whippany, New Jersey, USA) over 2 s. The power-contrast injector was synchronized to a fluoroscopy angiographic machine with a 1.2 s delay in the contrast injection. DSA images at a rate of 3 frames/s were acquired in anteroposterior, lateral, and oblique transorbital projections using a biplane neuroangiography suite (Artis zee, Siemens Healthineers, Erlangen, Germany). DSA images were routinely saved to the picture archiving and communication system (PACS) and the entire unedited DSA was archived on a separate DVD in DICOM (digital imaging and communication in medicine) format.
DSA images were downloaded onto a central computer for analysis. Individual DSA runs were analyzed for contrast image intensity throughout the angiogram cycle with a custom-made software code (MATLAB, Mathworks, Natick, Massachusetts, USA). This technique identifies various intensity plots, including the maximum gray intensity in consecutive DICOM images.12 ,13 Region of interest (ROI) was selected over the cavernous ICA ipsilateral to the AVM and over each of the AVM draining veins at their closest point to the nidus. Only lateral projections were used since they provided the best visualization of the cavernous ICA and draining vein. A-Vt was defined as the difference between peak contrast intensity in the cavernous ICA ipsilateral to the AVM and peak contrast intensity in the draining vein (figures 1 and 2). For AVMs with multiple draining veins, peak contrast intensity was measured within each draining vein and averaged. The A-Vt index was defined as the ratio of the draining vein diameter to A-Vt:
If multiple draining veins were present, then the A-Vt index was the sum of the ratio for each draining vein:
Contrast transit time to the AVM draining vein at its closest point to the nidus was determined using the commercially available postprocessing software syngo iFlow (Siemens Healthineers, Erlangen, Germany) (figure 2).9 ,10 For AVMs with multiple draining veins, contrast transit time to each draining vein was measured and then averaged.
AVM flow measurements
All patients underwent quantitative flow measurements of the extracranial and intracranial arteries and veins using QMRA. This technique of blood flow quantification by QMRA combined with Noninvasive Optimal Vessel Analysis (NOVA) software (VasSol, Inc, River Forest, Illinois, USA) has been described previously14 and validated using in vitro and in vivo models.15 It has been shown to be useful in the hemodynamic evaluation of cerebral AVMs and several other cerebrovascular diseases.16–25
Total AVM flow was measured from single draining veins where possible (n=12) or derived based on aggregate flow (n=21) within primary arterial feeders relative to flow in their contralateral counterparts according to the following equation:
Pearson's correlation was used to assess the relationship between A-Vt, iFlow, and AVM flow. Mean A-Vt, iFlow transit times, and AVM flow were compared in the presence and absence of AVM anatomical and clinical features using the Wilcoxon rank sum test. Analyses were performed with Stata (V.12.0, StataCorp, College Station, Texas, USA).
Thirty-three patients (17 male, 16 female) were included. Clinical, anatomical, and hemodynamic characteristics of AVMs in our cohort are summarized in table 1.
A-Vt and iFlow versus AVM flow
A-Vt (r=−0.47, p=0.01) and the A-Vt index (r=0.36, p=0.04) correlated significantly with total AVM flow measured using QMRA. iFlow transit time (r=−0.44, p=0.01) was also significantly associated with AVM flow (figure 3).
A-Vt, iFlow, and AVM flow versus AVM angioarchitecture and clinical presentation
Mean A-Vt did not significantly vary with AVM anatomical and clinical features, but the mean A-Vt index was significantly different among AVMs with a single draining vein (2.8 vs 10.5 mm/s, p<0.01) and an intranidal fistula (12.5 vs 6.6 mm/s, p=0.05). Mean iFlow transit times were not significantly different in the presence or absence of AVM anatomical and clinical features, except that iFlow transit time was significantly shorter in patients who presented with seizure (4.3 vs 5.7 s, p=0.02). Mean AVM flow was significantly lower with hemorrhagic presentation (145 vs 281 mL/min, p=0.04) and higher with seizure presentation (351 vs 197 mL/min, p=0.02). However, AVM flow did not vary with other angioarchitectural features. All results are shown in table 2.
The characterization of cerebral AVM hemodynamics is recognized as an essential component of the evaluation and management of AVMs. An analysis of AVM hemodynamics is particularly useful before and after embolization procedures and can guide treatment strategies by providing a hemorrhage risk assessment. Cerebral AVM flow has previously been characterized with four-dimensional flow MRI and time-resolved spin-labeled MRA.26–28 Additionally, our group examined AVM hemodynamics using QMRA before and after embolization and with respect to various angioarchitectural features thought to be associated with AVM rupture.16 ,17 ,21–25
Cerebral AVM flow has also been described and hemorrhage risk evaluated using DSA.5–10 For instance, Matsumoto et al5 reported shorter peak tracer transit times among ruptured AVMs, and similarly Todaka et al6 found significantly shorter mean contrast transit times within feeder arteries of ruptured AVMs. Currently, DSA assessment of AVM flow can be performed with commercially available software, such as syngo iFlow (Siemens Healthineers, Erlangen, Germany) and AngioViz (GE Healthcare, Chicago, USA).9 ,10 Importantly, DSA has a distinct advantage over MRA by enabling real-time acquisition of quantitative data in the neuroendovascular surgery suite.
However, AVM flows obtained using DSA have not been validated against true flow measurements, thereby hindering accurate interpretation of the results reported in the aforementioned studies. DSA determination of AVM hemodynamics also generally relies on a time to peak analysis that can be affected by vessel size, amount of contrast injected, and frame rate.12 ,29 Consequently, mainstay techniques are dependent on the location of the catheter during the contrast injection, vessel diameter, and volume of contrast injected, which may lead to variable and imprecise results.
Here, we introduce a new method of DSA image processing to determine cerebral AVM hemodynamics that is based on the ROI rather than time to peak. Our flow analysis, then, is not confounded by the site of the contrast injection, vessel size, or volume of contrast administered. Additionally, the A-Vt index was used to control for the draining vein diameter within each AVM. Our group previously described this contrast time–density technique in a subarachnoid hemorrhage model.12
Moreover, for the first time we attempt to validate cerebral AVM hemodynamics assessed by DSA using QMRA. Our results show that A-Vt (r=−0.47, p=0.01) and the A-Vt index (r=0.36, p=0.04) correlate significantly with total AVM flow measured by QMRA, demonstrating that these parameters attained from DSA accurately reflect true flow measurements. Despite the differing technique, we also found that iFlow transit times correlate significantly with QMRA flows (r=−0.44, p=0.01), establishing that iFlow estimates AVM hemodynamics equally as well as A-Vt. A-Vt and iFlow transit times, then, may be used to indirectly estimate AVM flow before and after embolization during angiography in real-time (figure 4).
We also examined the relationship between A-Vt, iFlow transit time, AVM angioarchitecture, and hemorrhage. We found that A-Vt and iFlow generally did not vary with AVM angioarchitectural characteristics or nidus volume, but the iFlow transit time was significantly shorter in patients who presented with seizure just as AVM flow was higher is these patients. The mean A-Vt index was significantly different among AVMs with a single draining vein and AVM flow was lower in ruptured AVMs, supporting the relative importance of outflow resistance over total inflow in AVM rupture risk. Based on these results, we further posit that the A-Vt index, as opposed to A-Vt and iFlow, is probably the best indicator of the presence of AVM anatomical features associated with elevated hemorrhage risk. Overall, these results are congruent with previously documented findings on the angiographic determinants of hemorrhage from cerebral AVMs.5 ,6 ,11 ,16
Possible limitations of this study include its retrospective design and small sample size, but it is the first study of its kind. Potential variability in the data stems from the fact that blood flow is a physiological parameter affected by age, heart rate, and blood pressure. Additionally, among patients with ruptured AVMs, QMRA and DSA were obtained immediately after hemorrhagic presentation, and so intraparenchymal hematoma with local mass effect might have affected accurate assessment of the hemodynamics. We attempted to mitigate inherent limitations of our DSA image processing technique by calculating the A-Vt index and selecting the ROI within the draining vein at its closest point to the nidus in order to reduce the washout effect from non-opacified blood or confounding contributions from physiological drainage. Nevertheless, this preliminary validation study uniquely assures that our contrast time–density analysis coincides with actual flow measurements. While our DSA assessment of AVM flow did not detect a difference in A-Vt or iFlow among AVMs with venous stenosis (n=6) or hemorrhagic presentation (n=9), our A-Vt index could distinguish between AVMs with and without a single draining vein (n=12 vs n=21, respectively). Additional studies with larger sample sizes may be better powered to discern the relationship between outflow obstruction and AVM rupture.
A-Vt and iFlow transit times on DSA correlate with cerebral AVM flow measured using QMRA. Thus, these parameters may be used to indirectly estimate AVM flow before and after embolization during angiography in real time.
Contributors SFS: drafted and revised manuscript. DB: data collection and statistical analysis. AEH and AL: data collection and review of the manuscript. C-YH: data collection. FTC: Reviewed rmanuscript. AA: devised and supervised the project. Critically revised manuscript.
Competing interests AA: research grant, NIH; consultant, Cordis-Codman. FTC: ownership interest, VasSol Inc; consultant, Transonic.
Ethics approval University of Illinois at Chicago.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.