Article Text

Download PDFPDF
Original research
The application of susceptibility-weighted MRI in pre-interventional evaluation of intracranial dural arteriovenous fistulas
  1. Narendra Kumar Jain,
  2. Santhosh Kumar Kannath,
  3. Tirur Raman Kapilamoorthy,
  4. Bejoy Thomas
  1. Department of Imaging sciences and Interventional Radiology, Sree Chitra Thirunal Institute of Medical Sciences and Technology, Trivandrum, Kerala, India
  1. Correspondence to Dr Bejoy Thomas, Department of Imaging Sciences and Interventional Radiology, Sree Chitra Thirunal Institute for Medical Sciences and Technology, Trivandrum 695011, Kerala, India; bejoy{at}


Background and purpose Detection and characterization of intracranial dural arteriovenous fistula (DAVF) is important to plan appropriate therapeutic management. The aim of this study was to analyze the utility of susceptibility-weighted MRI (SWI) in the pre-therapeutic assessment of DAVF in comparison with gold standard digital subtraction angiography (DSA).

Materials and methods Prospectively, 26 patients with DAVFs underwent a thorough clinical examination and MRI including SWI followed by cerebral DSA. Two observers blinded to the DSA findings evaluated conventional MRI and SWI images and identified the fistulous area (FA), cortical venous reflux (CVR), and cortical venous ectasia (CVE) and compared these observations with the DSA findings documented by a third observer.

Results Aggressive clinical symptoms were observed in 31% of patients and benign features were noted in 69% of DAVFs. Conventional MRI could identify the FA in only 27% of patients. SWI accurately located 75% of all the FAs in 23 patients. However, SWI failed to identify DAVFs in three patients. CVR was detected in 89.6% of all aggressive DAVFs. The accuracy of SWI to identify CVE was 100% and the extent and degree correlated with DSA observations.

Conclusions SWI is a reliable non-invasive tool for the localization and characterization of DAVFs and is superior to conventional MRI in the evaluation of DAVFs. This sequence can demonstrate underlying cerebral hemodynamic stresses with a high degree of accuracy and provide valuable pre-therapeutic information.

  • Fistula
  • Magnetic Resonance Angiography

Statistics from

Request Permissions

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.


Intracranial dural arteriovenous fistulas (DAVFs) are uncommon neurovascular diseases which account for approximately 10–15% of all intracranial arteriovenous lesions.1 Studies on the natural history of these lesions have noted the relationship of the clinical symptoms with patterns of venous drainage.2–4 Aggressive clinical presentations with increased risk of hemorrhagic as well as non-hemorrhagic neurological complications are seen with DAVFs with cortical venous reflux (CVR).5 ,6 CVR leads to venous congestion and causes functional venous obstruction leading to hypoperfusion. These mechanisms together can explain most of the neurologic complications.7 Hence, identification of CVR in the evaluation of DAVF is of paramount importance. Moreover, the management of these fistulas is primarily directed towards elimination of CVR and in radiological studies it is important to assess the presence of CVR.

Digital subtraction angiography (DSA) remains the gold standard for assessing dural fistulas including the presence of CVR. Although the incidence of serious complications of cerebral angiography is <1%, patients with DAVF often need multiple DSA evaluations in the pre- and post-therapeutic phases to diagnose, assess the adequacy of treatment, and detect recurrence.8 Non-invasive imaging, which improves the diagnostic accuracy, is therefore worthwhile in the evaluation of DAVF. Susceptibility-weighted MRI (SWI) is a relatively new technique that combines information from both phase and magnitude images to accentuate the detection of intravascular venous deoxygenated blood as well as extravascular blood products, and its applications in the evaluation of vascular and non-vascular pathologies have been widely reported.9–13

However, the literature on the use of SWI for DAVF evaluation is sparse and is limited to a few case reports or case series.14–21 The diffuse prominence of the venous vasculature noted on SWI is due to slow transit of blood in the intracranial circulation resulting from functional obstruction caused by DAVFs at the level of the venous sinuses or cortical veins. This allows for increased oxygen extraction, resulting in more desaturation of blood in the veins. The possible site of the fistula may also be inferred as an area of hyperintensity on magnitude images of SWI.19 ,20

The aim of this study was to assess the utility of SWI in a larger series of intracranial DAVFs with special emphasis on the evaluation of CVR non-invasively, keeping DSA as the gold standard.

Materials and methods

Patient characteristics

All consecutive patients referred to our institution with a diagnosis of intracranial DAVF from February 2013 to September 2014 were prospectively enrolled in the study. Patients with contraindications for undergoing MRI or DSA were excluded. The institute ethics committee approved the study and informed written consent was obtained from all the participants.

Study protocol

All patients underwent a detailed neurologic evaluation and the modified Rankin score was documented. The symptoms were categorized as (1) benign if the patient was asymptomatic or presented with headache, tinnitus, seizures or orbital symptoms; or (2) malignant if the patient presented with intracranial bleed, non-hemorrhagic neurologic deficits, or dementia. The subjects underwent initial MRI including SWI, followed by DSA within a week.

MRI and SWI techniques

MRI was performed on a 1.5 Tesla scanner (Magnetom Avanto; Siemens, Erlangen, Germany) with a 12-channel head coil. Conventional sequences such as T2-weighted imaging (T2WI), fluid-attenuated inversion recovery (FLAIR), diffusion-weighted imaging (DWI), and time-of-flight MR angiography (MRA) were routinely obtained in all patients. The cervical spine was evaluated if perimedullary venous drainage or prominent veins in the posterior fossa were evident or suspected. The SWI sequence was obtained in the axial plane with the following imaging parameters: TR, 49 ms; TE, 40 ms; number of averages, 1; flip angle, 15°; slice thickness, 2 mm; matrix size, 448×290; pixel bandwidth, 80 Hz; field of view (read), 230 mm; field of view (phase), 186 mm. Raw data processing was performed automatically using Magnetom Syngo software (Siemens) and yielded magnitude image, phase image, SWI image, and minimum intensity projections (mIP).

Digital subtraction angiography (DSA)

All angiograms were performed on an Innova biplane flat panel DSA unit (Innova 3100; GE, Milwaukee, USA). Diagnostic angiograms were performed under local anesthesia and selective angiograms of bilateral internal carotid arteries, external carotid arteries, and vertebral arteries were obtained. Superselective injections from the specific external carotid artery feeders were also obtained when deemed necessary.

Image analysis

All angiograms were evaluated by an interventional neuroradiologist (SKK with 5 years of experience) and the DAVFs were categorized according to the Borden and Cognard classifications (figures 1A, B and 2A, B). The angiographic features were tabulated for: (1) type of fistula; (2) location of fistula; (3) presence of CVR; and (4) cortical venous ectasia (CVE). SWI images were evaluated in consensus by two diagnostic neuroradiologists (BT and NKJ with 17 and 2 years of experience, respectively, who were aware of the presence of a DAVF but were blinded to the angiographic findings). The conventional MRI was analyzed and the location of the fistula was categorized into three groups (definite, probable or inconclusive) based on the emergence of draining veins from a major sinus or a prominent venous sac: ‘definite’ was defined when the draining veins could be seen arising from a major sinus or venous sac; ‘probable’ when the draining veins could be traced to the sinus or particular location but the actual site was not visualized; and ‘inconclusive’ when the site could not be identified by MRI. CVR and CVE were not assessed as it was difficult to ascertain whether the prominent veins observed were due to venous reflux or venous rerouting secondary to chronic venous thrombosis. SWI was assessed for the presence of CVE, site of the fistula, and the presence or absence of CVR. CVE on SWI was defined as an increase in the number, tortuosity, and caliber of leptomeningeal and/or medullary veins compared with that in the normal appearing regions of the brain (figures 1C and 2C, D). A fistulous area (FA) was defined as hyperintensity within a venous structure, either the cortical vein or venous sinus (figure 1D). Signal intensity equal to or greater than that of the proximal intracranial arteries at the circle of Willis was used to define venous hyperintensity. CVR was defined as hyperintensity in cortical venous structures radiating from the presumed FA (figure 2C). The identification of FAs and CVR were made in magnitude and SWI images. mIP images were used for assessing the CVE. SWI images were also assessed for the presence of evidence of recent or old intracranial hemorrhage.

Figure 1

A middle-aged patient presented with tinnitus in the left ear. (A, B) Digital subtraction imaging revealed a type IIa dural arteriovenous fistula located at the left transverse sigmoid sinus junction with cortical venous ectasia (CVE). (C) CVE was identified on susceptibility-weighted MRI (SWI) (arrow). (D) SWI identified the fistulous area at the left transverse sigmoid junction (arrow).

Figure 2

A patient presented with progressive diminution of vision, redness and pain in the eye. (A, B) Digital subtraction imaging revealed a type IIa+b dural arteriovenous fistula located at the mid superior sagittal sinus with significant cortical venous reflux. (C) Susceptibility-weighted MRI identified the fistulous area (arrow) at the superior sagittal sinus with the presence of cortical venous reflux (arrowhead) and cortical venous ectasia on minimum intensity projections (D).

Statistical analysis

Statistical analyses were performed using the Statistical Package for Social Sciences V.18 for Windows (SPSS, Chicago, Illinois, USA). The parameters evaluated on SWI such as FAs and the presence of CVR and CVE were qualitative in nature and were compared with the findings on the angiogram which was taken as the gold standard. Diagnostic accuracy indicators were calculated.


A total of 26 patients (22 men) aged 17–60 years (mean 44 years) with intracranial DAVF were included in the study. Eight patients (31%) had aggressive symptoms while the remaining 18 patients (69%) had non-aggressive symptoms. Demographic, clinical, MRI, and angiographic characteristics are shown in tables 1 and 2.

Table 1

Clinical and demographic data of patients with dural arteriovenous fistula (DAVF)

Table 2

Angiographic and SWI correlation of DAVFs

Angiographic findings

Six patients had multiple fistulas, of which one patient had three FAs and the remaining five had two FAs each on DSA. A total of 33 FAs were observed. The locations of these fistulas are shown in table 3. Of the 33 FAs evaluated on DSA, 29 were associated with CVR. However, only one patient had Borden type I DAVFs with no CVR and the rest of the patients had a single DAVF with CVR or multiple fistulas of both benign and aggressive types. Three patients demonstrated spinal perimedullary venous drainage. One patient with spinal perimedullary venous drainage had hyperintensity on T2WI in the upper cervical cord. CVE on DSA was identified as dilated tortuous collateral and bridging veins seen on the venous phase of the angiogram (figure 2B), associated with prolongation of circulation time. Angiograms of 15 patients (58%) revealed CVE while the remaining 11 patients did not have venous ectatic changes. Among the patients with CVE, six showed severe venous congestion with significant prolongation of circulation time. Milder changes were observed in the remaining nine patients.

Table 3

Identification of FA location on DSA and SWI

Conventional MRI

Conventional MRI identified flow voids in 46% of the patients. The FA could be definitively identified in only two patients and the probable location of the fistula was identified in 19% of patients. In the vast majority of patients (73%) the site of the fistula could not be characterized based on direct or indirect imaging signs. MRI and DSA showed 85% concurrence in positively diagnosed patients.

Susceptibility-weighted MRI (SWI)

The correlative analysis between DSA and SWI findings are presented in table 2. SWI identified at least one FA in 23 of the 26 patients (88.5%); however, only 25 of the 33 FAs (75%) could be accurately detected using SWI. There were three patients in whom SWI could not identify any DAVF. Significant susceptibility artifacts from prior intracranial hemorrhages obscured accurate localization in two patients. Both of these patients had a Borden type III fistula, one in the parietal region and the other in the posterior fossa. The third patient had a Borden type II fistula at the left transverse–sigmoid sinus junction with large saccular aneurysms in the feeding artery. Six patients had multiple FAs on angiography. Of these, SWI could identify all the FAs in only one patient. However, in all these patients SWI could identify at least one FA. The fistulas which were not identified on SWI were smaller in size with low flow while the other co-existing fistulas were larger and of high flow. The altered flow dynamics in the venous sinuses due to a high flow fistula may have impaired the visualization of other co-existing smaller fistulas. Among the patients with single fistulas, the sensitivity of SWI detection was 85%. There was concordance between the DSA findings and SWI as far as the location of the fistula was concerned and there were no false positives on SWI. SWI was able to identify CVR in 22 of the 25 patients with DAVF. CVR was not detected in three patients with aggressive DAVFs, all Borden type II. SWI was able to detect prominent hypointense veins in all 15 patients with CVE. The extent of CVE seen on angiograms correlated with the number, tortuosity, and caliber of hypointense veins seen on mIP of SWI by one-to-one comparison with DSA in all these patients. However, no grading of CVE on SWI was attempted in the present study.


Owing to variable clinical presentation, patients with a suspected DAVF are initially screened with non-invasive imaging such as CT or MRI prior to cerebral DSA. However, normal routine imaging does not exclude the presence of DAVF. Several authors have studied the usefulness of various advanced imaging techniques in the evaluation and characterization of DAVF. Among the findings described, the presence of asymmetric arteries on CT was found to be 86% sensitive in the identification of DAVF and prominent leptomeningeal vessels on CT/MRI were significantly observed in aggressive types of DAVFs.22 ,23 Time resolved CT angiography was found to agree with DSA in 90% of the patients and could appropriately categorize different DAVFs as per Borden’s classification.24 ,25 Advanced MR sequences have the potential to study anatomic, functional, and hemodynamic aspects of dural fistulas and have been reported in several small case series.20 ,21 ,26 ,27 The accuracy of time resolved three-dimensional contrast enhanced MRA in the identification of DAVFs was found to be 100% and its utility in the characterization of DAVFs was reported to be between 77% and 85%.26 Using a non-contrast four-dimensional MRA technique in a small group of patients, Jang et al27 demonstrated and classified DAVFs in 100% and 85% of patients, respectively. However, there were important pitfalls with these techniques, such as poor spatial resolution, lack of directionality information, and obscuration of signals by thrombosis.

The utility of SWI in the diagnosis of DAVF was initially recognized in a case report which identified prominent hypointense veins in the cerebral parenchyma that disappeared following treatment of the DAVF.14 Tsui et al15 observed hyperintensities within the draining venous sac or sinuses due to rapid shunting of oxygenated blood. Another larger series of vascular malformations showed that the presence of these high signals could reliably distinguish high flow vascular malformations from low flow lesions with sensitivity, specificity, and accuracy of 93%, 98%, and 96.3%, respectively with high inter-observer agreement.17 However, only five DAVFs were included in the final analysis of the study. Noguchi et al evaluated SWI in 10 patients with DAVFs with retrograde cortical venous drainage and demonstrated that SWI was superior to conventional MR sequences in the evaluation of cortical as well as medullary veins.16 Susceptibility-weighted angiography (SWAN) identified the refluxing veins of DAVFs as hyperintense structures and attributed this observation to time-of-flight effects.19

DAVF is characterized by two systems of parallel circulation: (1) oxyhemoglobin-dependent direct variable degree arteriovenous shunting into the dural venous system and/or cortical veins; and (2) normal and/or delayed cerebral circulation with conspicuous effects on cerebral venous drainage. The former certainly can affect the latter normal cerebral parenchymal blood flow and result in variable clinical manifestations.28 SWI is a robust sequence that provides both anatomic and hemodynamic information and hence the sequence is best suited to study the different pathophysiologic aspects of DAVF.20 ,21 Cardinal imaging findings on SWI include hyperintense venous structures and the presence of prominent, dilated, and asymmetric cerebral veins in the cerebral hemisphere. Hyperintense veins help in the localization of the DAVF, and this observation is presumed to be due to the unsaturated and oxyhemoglobin rich blood spins shunting through the fistula which accentuates the SWI signal due to time-of-flight and diamagnetic effects.17 ,19–21 Hypointense prominent veins are correlated with CVE observed in DSA, and this has been attributed to increased oxygen extraction and a resultant higher concentration of deoxyhemoglobin secondary to prolonged transit of normal cerebral venous drainage.14 ,17 ,20

There have been only two reports in the literature that discuss the application of SWI in the characterization and classification of DAVFs. In a small series of six patients, Letourneau-Guillon et al correctly localized DAVFs in 83% of patients. CVR was identified in 60% of the patients with DAVF, but the extent was significantly underestimated compared with DSA. Good correlation was noted between the prominent veins noted by SWI and CVE observed by DSA.20 Nakagawa et al evaluated the efficacy of SWI in a series of 17 patients with aggressive DAVF with CVR in pre- and post-therapeutic phases. They observed that CVR was identifiable in 88% of patients and the finding was falsely negative in two patients, presumably due to slow flow. Venous congestive features such as prominent parenchymal veins were identified in 65% of patients. Additionally, successful obliteration of the fistula resulted in disappearance of hyperintense signals and temporally variable normalization of congested veins were also demonstrated in this study.21 These two studies had limitations such as retrospective methodology, small sample size, and variable imaging protocols. They did not evaluate the localization capability of SWI in DAVF.

In our study the sensitivity of SWI in localizing the DAVF was 85%, which was reduced to 75% when multiple fistulas were included. The location of the fistula corresponded with DSA in all detected cases and there were no false positive observations with SWI. This suggests that SWI is a reliable tool in the identification of the fistula location. Compared with DSA, CVE was detected in all positive cases on SWI; however, CVR was observed in only 89% of the fistulas. Prior parenchymal hemorrhage impaired the identification of DAVF and subtle cortical reflux or turbulent flow limited CVR detection. In contrast, conventional MRI was found to be inferior in the diagnosis and characterization of suspected DAVF.

Our findings are comparable with other previous reports on comparable MRI techniques.20 ,21 ,24 ,26 Although localization is superior with contrast-based MR techniques, SWI has the distinct advantage of being a non-contrast technique which provides additional information on the underlying hemodynamic stress.

The present study is the largest reported series to date to assess the role of SWI in the evaluation of DAVFs. The data were prospectively acquired with uniform protocols and both benign and aggressive DAVFs were included. There were some limitations of our study. Only patients with a proven diagnosis of DAVF were included in the study analysis and hence the investigators were aware of the presence of a fistula and may have been biased towards careful search for identification of FA. The specificity of SWI in DAVF evaluation could not be calculated because of the inclusion criteria. The present study localized the DAVFs based on their anatomical location and the relation to sinus or cortical veins and could not classify the DAVFs based on the angiographic grading systems due to the lack of directionality information of venous flow.


Although cerebral DSA remains the gold standard investigation for DAVFs, SWI sequence helps in the pre-therapeutic characterization of DAVFs. SWI is reliable in locating DAVFs and can detect CVR and CVE with improved sensitivity. Inclusion of this sequence in the MR protocol provides unique information regarding the morphological and hemodynamic characteristics of the DAVF and thus improves the diagnostic accuracy of MRI.



  • NKJ and SKK contributed equally.

  • Contributors NKJ: acquisition of data, data analysis, manuscript preparation. SKK: study design, data analysis, manuscript preparation, critical revision of manuscript. TRK: critical revision and guarantor of the study. BT: study design, manuscript preparation, critical revision and study guarantor. All authors reviewed the manuscript.

  • Competing interests None declared.

  • Ethics approval Institutional ethics committee.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement The authors agree to share any data on request to SKK (