Vessel wall imaging and quantitative flow assessment in arteriovenous malformations: A feasibility study

Abstract

Introduction

Cerebral arteriovenous malformations (AVMs) carry a rupture rate of 2–3% per year. Several architectural factors may influence rupture rate, and a recently theorized model of AVMs describes the influence of vessel wall inflammation. A novel imaging modality, vessel wall imaging (VWI), has been developed to view inflammatory processes in vessel wall foci but has not yet been examined in AVMs, which is the aim of this study.

Methods

This retrospective review studies prospectively collected data on patients with ruptured and unruptured AVMs between 2019 and 2021. Inclusion criteria included adult patients (≥18 years) with radiographically diagnosed AVM who underwent VWI. Charts were reviewed for medical history, clinical presentation, hospital course, discharge condition, and follow-up. Angioarchitectural features, blood flow, and VWI were compared in patients with and without hemorrhagic patients.

Results

Nine patients underwent VWI, mean age 37.7 ± 9.9 years. Four presented with hemorrhage (44.4%). Seven (77.7%) received glue embolization and 6 (66.7%) underwent surgical resection. All patients (4/4) with a history of hypertension presented with hemorrhage (p = 0.0027). Size and Spetzler-Martin grade were not associated with hemorrhage (p = 0.47, p = 0.59). Net AVM flow was higher in patients presenting with hemorrhage, although nonsignificant (p = 0.19). With VWI, 3 (75%) hemorrhagic AVMs showed visible nidus and draining veins, and all three demonstrated positive post-contrast wall enhancement in at least one of their draining veins; conversely, of fivenonhemorrhagic AVMs, only 2 (40%) demonstrated post-contrast wall enhancement in any draining vein (p = 0.090).

Conclusion

This pilot study successfully demonstrated capture of venous walls in AVMs using VWI. In this study, draining vein enhancement occurred more often in hemorrhagic AVM and in those with higher venous volumetric flow.

Introduction

Cerebral arteriovenous malformations (AVMs) carry a rupture rate of 2–3% per year, which amounts to a 75% cumulative lifetime rupture risk for a newly diagnosed patient at the age of 30 years.^1,2 Several architectural factors may influence rupture rate, including size of AVM, deep venous location, hypertension, and reduced venous output.^3,4 A recently theorized model of AVMs describes the influence of vessel wall inflammation that predisposes to stenosis and, ultimately, hemorrhage of the nidus. ⁵

A novel imaging modality, termed vessel wall imaging (VWI), has been developed to study inflammatory processes in vessel wall foci. VWI utilizes T1-weighted magnetic resonance imaging (MRI) with fluid-attenuated inversion recovery (FLAIR) sequencing to suppress cerebrospinal fluid. ⁶ Prior applications of VWI in cerebral aneurysms have demonstrated that wall thickness and the degree of macrophage infiltration correlate with the prevalence of contrast enhancement ⁷ and have shown that symptomatic aneurysms exhibit a greater prevalence of post-contrast wall enhancement relative to non-symptomatic aneurysms. ⁸ Further, aneurysms with certain features that predispose to hemorrhage, such as including higher PHASES scores, greater volume, and greater height, demonstrate a higher prevalence of wall enhancement.^9,10

VWI technology has not yet been examined in AVMs. Although VWI has been previously applied to other forms of cerebrovascular pathology, the dynamic and tortuous nature of AVMs poses unique challenges with respect to wall imaging. This pilot study serves as a proof-of-concept investigation into the feasibility of VWI application to AVMs and aims to determine if VWI can detect the vessel wall, assess the venous flow using quantitative magnetic resonance angiography (QMRA), and evaluate the contribution of these factors in hemorrhagic versus nonhemorrhagic AVMs.

Methods

This retrospective review studies prospectively collected data on patients with ruptured and unruptured AVMs. Angioarchitectural features, blood flow, and results of VWI were compared in patients with and without hemorrhagic patients using the following parameters.

Patient selection

This retrospective chart review examines patients 18 years and older who presented to this institution with a radiographically diagnosed AVM. These patients underwent VWI as part of their clinical care from October 2019 to January 2021. Data were collected prospectively as part of a quality improvement database at this institution. Institutional review board approval was obtained with a waiver of patient consent in the setting of retrospective study, compliant with HIPAA. Patients who could not undergo VWI due to a contraindication to MRI or gadolinium contrast were excluded. Charts were reviewed for medical history, including clinical presentation, hospital course, discharge condition, and follow-up.

Imaging

VWI images were acquired on a standard 32-coil 3 Tesla MRI (Phillips Ingenia®; Cambridge, MA, USA). VWI consisted of T1-weighted spin-echo FLAIR, with and without contrast, as described in the literature,^6,9,11 which generated two-dimensional image outputs in three orthogonal planes. To determine hemorrhagic status and AVM location, non-contrast computed tomography (CT) and MRI with T2-weighted FLAIR were performed. Digital subtraction angiography (DSA) was used to determine AVM architecture, feeder arteries, and draining veins. To quantify vascular flow, QMRA was completed using Non-invasive Optimized Vessel Analysis (NOVA; VasSol, Inc.; River Forest, IL, USA). Imaging studies were review by a trained neuroradiologist and neurosurgeon.

QMRA was reviewed to determine the vascular flow per draining vein (when available) and net AVM flow using two general methods. First, in cases in which draining veins were successfully captured by QMRA imaging, the sum of flows through the draining veins was calculated as net AVM flow. ¹² Otherwise, in cases in which draining vein flows were not captured, a formula calculating ipsilateral net flow excess was employed:

$$NetAVMFlow=Ipsilateral(A1+M1+PCA)\text{–}Contralateral(A1+M1+PCA),$$

where “A1” represents the first segment of the anterior cerebral artery, “M1” represents the first segment of the middle cerebral artery, and “PCA” represents the posterior cerebral artery. ¹²

VWI was reviewed to determine the presence of post-contrast enhancement per draining vein. Draining veins identified on angiography were followed on axial, coronal, and sagittal images from their point of origin from the nidus to their juncture with their respective sinus. “Positive enhancement” was defined qualitatively as a post-contrast signal exceeding pre-contrast signal in the wall of a draining vein anywhere along its course from nidus to sinus. Of note, lumen-based enhancement or elevated signal on pre-contrast imaging was not counted as positive. “Circumferential enhancement” was defined as enhancement throughout the entire circumference of the draining vein wall on a transverse cut, or evident on at least two opposing wall sides (i.e. superior and inferior) on a tangential cut.

Statistical analysis

Chi-squared test was run to compare the prevalence of various demographic and radiographic factors between hemorrhagic versus nonhemorrhagic groups, as well as between enhancing versus non-enhancing draining veins. Student's t-test assuming unequal variances was run to compare quantitative measures between hemorrhagic versus non-hemorrhagic groups, as well as between enhancing versus non-enhancing draining veins. Given the small sample size (n = 9), multivariate analyses were not performed due to few degrees of freedom and low power of the analysis. The significance value was set at an alpha of 0.05. Statistical analysis was performed using SAS Studio® software (Cary, NC, USA).

Results

Baseline patient characteristics

A total of nine patients underwent VWI. A total of 4 (44.4%) patients exhibited hemorrhage on presentation. Three patients (33.3%) identified as African American; 4 (44.4%) identified as Hispanic; and 2 (22.2%) identified as White, non-Hispanic. The patient’s age ranged from 22 to 56 years, with a mean age of 37.7 ± 9.9 years. Four (44.4%) patients had no significant medical history; 4 (44.4%) had hypertension, 1 (11.1%) was diabetic, and 2 (22.2%) were current/former smokers. Six (66.7%) patients presented with seizure, 4 (44.4%) presented with headache, 2 (22.2%) with visual complaints, and 2 (22.2%) with motor weakness. Seven (77.7%) received glue embolization, and 6 (66.7%) underwent surgical resection. Three patients (33.3%) returned to baseline exam upon discharge, with the remainder (66.7%) demonstrating some residual neurologic deficit. Table 1 summarizes demographic and clinical presentation data. Of these factors, only a history of hypertension was associated with hemorrhage (p = 0.0027).

Table 1.

Baseline patient demographic and clinical data.

ID No.	ICH	Age/Sex	PMH	Present	Embo	No. of Embo	Surgery versus SRS	Outcome
1	Y	37/M	HTN, tobacco	Seizure, HA	Y a	1 a	SRS	Return to baseline exam
2	N	40/M	None	L weak	N	NA	None	L upper and lower extremity weak
3	N	34/M	None	Visual changes	Y	3	Surgery	Return to baseline exam
4	N	30/M	Non	Seizure	Y	4	Surgery	R upper/lower extremity weak, R facial droop
5	Y	56/F	HTN, MS	R weak, R numb	Y	1	Surgery	Return to baseline exam
6	N	33/F	HTN, PE, tobacco, marijuana	Seizure, HA	Y	8	Surgery	R temporal field visual deficit, persistent seizures
7	Y	39/M	HTN, DM	Seizure, HA, L temporal field cut, blurry vision	Y	2	Surgery	L upper extremity weak, L facial droop
8	N	48/F	None	Seizure	Y	4	Surgery	L upper quadrantanopia, transient paresthesia (possible TIA), persistent seizures
9	Y	22/F	Depression, seizure	Seizure, HA	Y a	1 a	None	Dysarthria, R upper/lower extremity weakness, R facial droop

^a Embolization performed aneurysm but not nidus.

DM: diabetes mellitus; HA: headache; HTN: hypertension; ICH: intracerebral hemorrhage; L: center; N: no; MS: multiple sclerosis; PE: pulmonary embolism; SRS: stereotactic radiosurgery; TIA: transient ischemic attack; Y: yes.

Radiography and architecture

The locations of AVMs varied: 2 (22.2%) were frontal, 2 (22.2%) occipital, 3 (33.3%) parietal, 1 (11.1%) temporal, and 1 (11.1%) deep. Size also ranged considerably: 2 (22.2%) had small AVMs (<3 cm in the longest dimension); 6 (66.7%) had mid-sized AVMs (<6 cm in the longest dimension); 1 (11.1%) had a large AVM (≥6 cm in the longest dimension). Similarly, 2 (22.2%) patients had low Spetzler-Martin grades (scores: 1–2), 4 (44.4%) had mid-level grades (score: 3), and 3 (33.3%) had high grades (scores: 4–5). 4 of 9 (44.4%) patients presented with hemorrhage. One patient had stenosis of the draining vein. Feeder artery, draining vein, and flow data are summarized in Table 2.

Table 2.

Radiographic, architectural, and vessel wall imaging (VWI) data.

ID No.	ICH	Location	Size	SM grade	Feeder arteries	DV number	DV site	Enhance.	Circum. Enhance.	Flow DV	Net AVM flow
1	Y	Posterior incisural	3 × 2 cm	3	Bilateral SCA, bilateral PCA, basilar perforator	1	Deep to straight sinus	Y	Y	105	105
2	N	Right frontal	5.5 × 5 cm	4	R ACA, R MCA	2	Superficial to SSS	Y	Y	430	604
							Deep to straight sinus	Y	Y	174
3	N	Left occipital	4 × 3.2 cm	3	L PCA	1	Superficial to TS	N	N	Not measured	508 a
4	N	Left frontal	4.3 × 4 cm	3	L ACA, L MCA	3	Anterior to SSS	Y	Y	156	337
							Posterior to SSS	N	N	33
							Inferior to TS	Y	Y	148
5	Y	Left parietal	2 × 2 cm	2	L MCA	1	Superficial to SSS	NA	NA	NA	NA
6	N	Right parietal	5.5 × 5 cm	2	R ACA, R MCA, R PCA, MMA	1	Superficial to SSS	N	N	778	778
7	Y	Right occipital	4.5 × 4 cm	4	R ACA, R MCA, R PCA, and L occipital artery	2	Deep to VoG	Y	Y	47.5	133.5
							Superficial to SSS + stenosis	Y	Y	86
8	N	Right temporal	3.5 × 3 cm	3	R MCA, R PCA	1	Deep to straight sinus	N	N	763	763
9	Y	Left parietal	6 × 6 cm	5	L ACA, L MCA, L PCA, and L ECA	3	Deep to VoG	N	N	Not measured	730 a
							Superficial, posterior to SSS	Y	N	Not measured
							Superficial, anterior to SSS	N	N	Not measured

^a Calculated with sum of ipsilateral A1, M1, and PCA minus the sum of contralateral A1, M1, and PCA.

AVM: arteriovenous malformation; Circum: circumferential; DV: draining vein; Enhance: enhancement; ICH: intracerebral hemorrhage; L: left; N: no; NA: not available; R: right; SM: Spetzler-Martin; SS: straight sinus; SSS: superior sagittal sinus; TS: transverse sinus; VoG: vein of Galen; Y: yes.

Although net flow was higher in patients presenting with hemorrhage, a Sudent's t-test assuming unequal variances demonstrated a nonsignificant association (p = 0.19). Additionally, size and Spetzler-Martin grade carried no association with hemorrhagic status (p = 0.47, p = 0.59, respectively).

VMI

There were three hemorrhagic AVMs with a visible nidus and draining veins (excluding case 5, with a non-visible nidus and draining vein). All three demonstrated positive post-contrast wall enhancement in at least one of their draining veins. Conversely, of the five nonhemorrhagic AVMs, only 2 (40%) demonstrated post-contrast wall enhancement in their draining veins. Thus, there is a trend toward preferential enhancement in the walls of draining veins of hemorrhagic AVMs relative to nonhemorrhagic AVMs, but not statistically significant (p = 0.090). Similarly, 2 of 3 (66.7%) hemorrhagic AVMs demonstrated circumferential enhancement in at least one draining vein, compared to only 2 of 5 (40.0%) nonhemorrhagic AVMs. Thus, there is a trend toward circumferential enhancement in the walls of draining veins of hemorrhagic AVMs relative to nonhemorrhagic AVMs, although nonsignificant (p = 0.47).

On a per-vein basis, there were a total of 14 identified draining veins among the eight patients (again, excluding case 5 which had no visualized draining vein). Table 2 demonstrates 10 draining veins which had luminal vessel flow measurements per QMRA. Student's t-test assuming unequal variances shows a nearly significant (p = 0.060) positive relationship between increased flow and draining vein enhancement, on a per-vein basis, and likewise between increased flow and circumferential venous enhancement (p = 0.060).

Case examples

Case 1: Example of ruptured AVM. A 37-year-old male with history of hypertension presented with seizures and loss of consciousness. CT demonstrated subarachnoid hemorrhage (Figure 1a). Angiogram identified 3 × 1.5 cm posterior incisural AVM fed by bilateral PCAs, superior cerebellar, and basilar perforators and draining into single deep vein that emptied into the straight sinus (Figure 1a). The right PCA feeder included an aneurysm, which underwent glue embolization. QMRA showed straight sinus flow of 205 cc/min (Figure 1c). Based on in-house data, we approximate native straight sinus flow to be 100 cc/min, and therefore, the AVM flow is approximately 105 cc/min. Pre-contrast VWI found identifiable draining veins with delineated walls, evident with multiple planes of view (Figure 1d and 1f); post-contrast images demonstrated circumferential enhancement of the draining vein (Figure 1e and 1g). The patient required ventriculoperitoneal shunt placement and returned to baseline status. He was discharged home in stable condition, with later completion of stereotactic radiosurgery.

Figure 1.

(A) Computed tomography (CT) without contrast demonstrates subarachnoid hemorrhage of the basal cisterns. (B) Angiography demonstrates a 3 × 1.5-cm arteriovenous malformation (AVM) near the splenium, with an evident single deep draining vein into the straight sinus. (C) Non-invasive Optimized Vessel Analysis (NOVA) imaging reveals a straight sinus flow of about 205 cc/min. Pre-contrast sagittal (D) and coronal (f) images demonstrate identifiable walls of the draining vein. (e and g) Post-contrast images demonstrate circumferential enhancement of the draining vein.

Case 8: Example of non-ruptured AVM. A 48-year-old female with epilepsy presented with worsening daily focal seizures. T2 FLAIR MRI revealed non-hemorrhagic right temporal AVM. Angiography characterized a 3.5 × 3-cm right temporal AVM with right MCA and PCA feeders and a single dilated deep vein draining to the straight sinus (Figure 2a and 2b). QMRA showed 763 cc/min flow through the single draining vein (Figure 2c). VWI demonstrated an identifiable ectatic draining vein with delineated walls (Figure 2d); however, post-contrast imaging revealed no draining vein wall enhancement (Figure 2e). The patient underwent embolization followed by surgical resection. Postoperatively, the patient exhibited a left upper quandrantanopsia but otherwise had baseline neurologic status. She had no recurrence of her daily seizure activity and has been maintained on daily antiepileptic therapy.

Figure 2.

(A and B) Angiography demonstrates a 3.5 × 3-cm arteriovenous malformation (AVM) with right ACA, MCA, and PCA feeders and a single deep draining vein to the straight sinus. (C) Non-invasive Optimized Vessel Analysis (NOVA) measurements demonstrate a flow of 763 cc/min through the single deep draining vein. Vessel wall imaging (VWI) imaging (D) demonstrates an identifiable ectatic draining vein with delineated walls coursing to the sinus. (E) Post-contrast VWI demonstrates a degree of intra-luminal enhancement but no wall enhancement on the post-contrast image.

Discussion

A recently theorized model of AVMs describes the influence of vessel wall inflammation that predisposes to stenosis and, ultimately, hemorrhage of the nidus. ⁵ Draining vein stenosis is an established risk factor for nidal hemorrhage.^4,13,14 Draining veins encounter abnormally high volumetric flows and turbulence, combined with their lack of a tunica media, makes them particularly susceptible to inflammation and stenosis. ⁵ This model is consistent with prior models of stenosis in a similar high-flow system: arteriovenous fistulas used for chronic renal dialysis, in which it has been demonstrated that the high turbulence of the fistular–vein junction predisposes to stenosis at that site.^15,16 The rapid transition from a high-flow, high-turbulence environment in the draining vein to the laminar flow typical of the sinus may similarly predispose to stenosis. ⁵ Notably, angiographic estimates of volumetric blood flow have demonstrated higher flow rates to be associated with stenosis, further validating the model of high turbulence as a contributor to stenosis. ¹⁷ Furthermore, histologic analysis of AVMs have demonstrated neutrophil and macrophage accumulation in vascular walls, and high-function variants of inflammatory genes have been associated with a greater risk of AVM hemorrhage. ¹⁸ Additional histologic analysis of stenotic draining veins have shown intimal hyperplasia, which correlates to the degree of blood flow within it. ¹² However, imaging vessel wall inflammation has historically been a challenge.

VWI has been developed to study inflammatory processes in vessel wall foci. VWI utilizes T1-weighted FLAIR MRI sequencing to suppress cerebrospinal fluid. ⁶ A high magnetic field strength 3 Tesla scanner and small voxel size (on the order of 2.0 × 0.4 × 0.4 mm) allow for high spatial resolution, and given a typical cerebral artery wall thickness of 0.2–0.3 mm, this small voxel size allows for evaluation of the wall. ¹⁹ Spin-echo pulsing allows suppression of dynamic components of the target area, thus effectively suppressing luminal blood and other contents. ¹¹ Gadolinium contrast enhancement also helps to visualize the vascular wall and characterize “static” inflammatory processes in the wall.^6,11 With regard to intimal hyperplasia and atherosclerosis, growth of vasa vasorum in the tunica media also allows for post-gadolinium contrast enhancement; in cases of vasculitis, the active inflammatory process of the vessel wall, with “leaky” membranes, is detected as post-contrast enhancement. ²⁰ Atherosclerosis demonstrates “eccentric” focal enhancement reflective of the focal nature of the plaque, whereas vasculitis demonstrates concentric enhancement reflective of the nature of the inflammation. ²¹ VWI may be used clinically to differentiate between atherosclerosis, vasculitis, and vasospastic etiologies of cerebrovascular pathology and in the diagnostic workup of stroke. ¹⁹

This pilot study shows the first application of VWI to cerebral AVMs. Importantly, draining veins were successfully characterized in VWI sequences. More specifically, the high-resolution 3T image allowed successful capture of the draining veins with delineated wall limits. Flow-suppressing spin-echo sequences allowed for better delineation of venous wall anatomy. Post-contrast enhancement was visualized in 5 of 8 (62.5%) of patients’ venous walls. Thus, this pilot study illustrates that VWI is capable of capturing the tortuous venous anatomy of AVM draining veins and allowing selective enhancement of vascular walls. Whereas prior applications of VWI have focused on three-dimensionally simple structures, such as aneurysms, this proof of concept demonstrates successful application of this technology to three-dimensionally complex structures.

All hemorrhagic AVMs (n = 3) with available imaging in this pilot study demonstrated positive circumferential post-contrast enhancement, relative to only 40% of nonhemorrhagic AVMs. Although the study is limited by its small sample size, this finding holds promise for the application of VWI technology to the prognostication of AVM stability. This study's findings echo prior studies in cerebral aneurysms, which have demonstrated preferential enhancement in aneurysms with higher risk factors for rupture, including those with higher PHASES scores, larger volume, and higher aneurysmal height.^9,10 One factor that should be noted is that hemorrhagic AVMs in the current study presented at the point of rupture. There is no scoring system for hemorrhage risk in AVMs comparable to the PHASES score in aneurysms, for example. Thus, the issue of prognosticating risk prospectively remains the work for a future longitudinal study.

Turbulence is thought to be a major contributor toward the intimal hyperplastic processes that contribute to venous stenosis. ⁵ Similar to arteriovenous fistulas,^15,16 AVMs bear a low-resistance conduit from arteries to veins, the latter of which are ill equipped to deal with high turbulence. In fact, high flow AVMs have been shown to bear a greater degree of venous stenosis on angiography. ¹⁷ It is important then that this study found a nonsignificant trend (p = 0.060) between venous flow and wall enhancement (n = 10). Thus, it is postulated that venous wall enhancement detected on VWI in fact measures active inflammatory processes triggered by the high-flow, turbulent environments of AVMs. Furthermore, these inflammatory processes may lead to sequelae, such as venous stenosis and increased propensity for rupture. Several inflammatory processes have been previously studied, including matrix metalloproteinase-9 and interleukin-6, and may be involved in this pathogenesis.^22–24

This study carries several limitations. As a pilot study, this technique requires further validation in AVMs. Furthermore, the dynamicity of VWI is not clear in AVMs, and changes in the type of enhancement may occur depending on prior rupture, embolization, or other yet unknown factors. Longitudinal studies should be performed to assess these potential variations in VWI over time and potentially relate this imaging finding to rupture risk, angioarchitectural changes (e.g. development of venous stenosis), treatment response, or more. As a small study, the results of the statistical analyses should be interpreted with caution, and a larger study would be required to confirm the differences seen between the two groups observed. In addition, histology of resected AVMs in the setting of VWI was unavailable for this pilot study but could add valuable insight.

Conclusion

This pilot study successfully demonstrated capture of venous walls in AVMs using VWI technology. In this study, draining vein enhancement occurred more often in hemorrhagic AVM and in those with higher venous volumetric flow. Further larger, prospective studies would be required to elucidate these trends and validate these findings.