Thin-Section MR Imaging for Carotid Cavernous Fistula

BACKGROUND AND PURPOSE: Carotid-cavernous ﬁ stulas are abnormal vascular shunts that can cause various neurologic or orbital symptoms. The purpose of this retrospective study was to evaluate the diagnostic performance of thin-section MR imaging for carotid cavernous ﬁ stula in patients with clinically suspected carotid cavernous ﬁ stula, and to identify possible imaging predictors of carotid cavernous ﬁ stula. MATERIALS AND METHODS: A total of 98 patients who were clinically suspected of having carotid cavernous ﬁ stula (according to their symptoms and physical examinations) between January 2006 and September 2018 were included in this study. The patients underwent pretreatment thin-section MR imaging and DSA. Thin-section MR imaging consisted of 2D coronal T1- and T2WI with 3-mm thickness and 3D contrast-enhanced T1WI with 0.6mm thickness. The diagnostic performance of thin-section MR imaging for carotid cavernous ﬁ stula was evaluated with the reference standard of DSA. Univariate logistic regression analysis was performed to determine possible imaging predictors of carotid cavernous ﬁ stula. RESULTS: Among the 98 patients, DSA con ﬁ rmed 38 as having carotid cavernous ﬁ stula. The overall accuracy, sensitivity, and speci-ﬁ city of thin-section MR imaging were 88.8%, 97.4%, and 83.3%, respectively. Possible imaging predictors on thin-section MR imaging included abnormal contour of the cavernous sinus (OR: 21.7), internal signal void of the cavernous sinus (OR: 15.3), prominent venous drainage ﬂ ow (OR: 54.0), and orbital/periorbital soft tissue swelling (OR: 40.4). CONCLUSIONS: Thin-section MR imaging provides high diagnostic performance and possible imaging predictors of carotid cavernous ﬁ stula in patients with clinically suspected carotid cavernous ﬁ stula. Thin-section MR imaging protocols could help decide appropriate management plans for patients with clinically suspected carotid cavernous ﬁ stula.

C arotid cavernous fistulas (CCFs) are abnormal vascular shunts connecting the carotid artery and cavernous sinus. The flow can originate either directly from the cavernous segment of the internal carotid artery (direct CCF), or indirectly through dural branches from the internal carotid artery or external carotid artery (indirect CCF or cavernous sinus dural arteriovenous fistula). 1 Common clinical symptoms of CCF include proptosis, conjunctival injection, chemosis, diplopia, headaches, subjective bruit, visual disturbance, and ocular pain. [2][3][4][5] The reference standard for the diagnosis and classification of CCF is DSA, which is also essential for management planning. 6,7 However, DSA is not routinely performed in all patients with suspected symptoms because of its semi-invasiveness. So, after consideration of clinical symptoms and imaging findings of CT, CTA, MR imaging, or MRA, physicians decide whether to perform DSA. DSA also presents limitations in the evaluation of structures other than vessels, such as orbital soft tissue and brain parenchyma.
CT, CTA, MR imaging, or MRA are usually used as the initial imaging technique for CCF. 8 The imaging findings on CT include proptosis, dilated superior ophthalmic veins, enlargement of extraocular muscles, orbital and periorbital soft tissue swelling, and focal bulging or diffuse distention of the cavernous sinus. Most CT findings are also visible on MR imaging. 9 -18 Additionally, an abnormal signal void within the cavernous sinus due to increased flow may also be visible in patients with CCF, though some studies described considerable numbers of false-positive and false-negative cases for this finding. 10,13 On MRA or CTA, visualization of flow signal or early contrast enhancement within the cavernous sinuses along with dilated draining veins can also be helpful for screening or diagnosing CCF. [19][20][21][22][23][24] Three-dimensional TOF MRA is known to be very sensitive in detecting CCF with some downside in the specificity. 19,22,25 More recent techniques to visualize CCF include 4D-CTA and arterial spin-labeled MR imaging. 26,27 With advances in MR imaging techniques, the resolution of MR imaging has notably increased over time and continues to do so. In head and neck imaging specifically, the fine structures of the cranial nerves and vascular anatomy have become more vivid owing to the introduction of thin-section MR imaging. [28][29][30] Due to its complex anatomy and relatively small size, the cavernous sinus is better depicted on thin-section MR imaging. 31 Additionally, subtle change of contour or signal change of cavernous sinus might be easily depicted on thin-section MR imaging. In the clinical setting, certain types of thin-section MR imaging for evaluating cranial nerve and orbit, are frequently used as the initial imaging technique for patients with orbital or cranial nerve symptoms. To the best of our knowledge, there is no published research on the diagnostic performance of thin-section MR imaging for CCF. Therefore, the purpose of our study was to evaluate the usefulness of thin-section MR imaging on a large cohort of patients with clinically suspected CCF.

MATERIALS AND METHODS
This retrospective study was approved by the review boards of our institution, and the requirement for informed consent for data evaluation was waived. The methods and reporting of the results are in accordance with the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) statement. 32 There was no external financial support for this study.

Study Population
The study population was obtained from a historical cohort of consecutive patients with clinically suspected CCF according to their symptoms and physical examinations, and who underwent pretreatment thin-section MR imaging and DSA between January 2006 and September 2018 at Asan Medical Center (2700 beds), an academic tertiary referral hospital in Seoul, Korea. The patients' symptoms included diplopia, eyeball pain, facial pain, ptosis, proptosis, periorbital swelling, eyeball injection, and visual disturbance. Patients were excluded from the study population if 1) they were under the age of 18, 2) the interval between thin-section MR imaging and DSA was more than 90 days, 3) thin-section MR imaging or DSA were obtained after treatment, and 4) the thin-section MR imaging or DSA image quality was poor. The patient selection procedure is summarized in Fig 1. Finally, 98 patients (32 men and 66 women) were enrolled. The age range of the patients was from 20 to 85 years, with a mean age of 54.6 years (standard deviation: 14.6 years).

Imaging Protocols
All patients underwent pretreatment thin-section MR imaging and DSA for evaluation of CCF. Because of the long study period, various MR imaging and DSA systems were used; however, most MR examinations were performed with a 3-T MR unit (Intera Achieva; Philips Medical Systems) by using a 16-element phased-array head coil. Two-dimensional coronal T1-weighted, T2-weighted (with and without fat suppression), and contrastenhanced T1-weighted turbo spin-echo images were obtained at a 3-mm thickness. Contrast-enhanced 3D T1-turbo field echo images were obtained with a 0.6-mm thickness and reconstruction section thickness was 1.2 mm with an interslice distance of 0.6-mm. The scan range of the coronal and axial images covered the orbit and cavernous sinus. Detailed parameters for the thinsection MR imaging are summarized in On-line Table. Cerebral DSA images included both internal carotid arteriograms, both external carotid arteriograms, and both vertebral arteriograms.

Interpretation of Images
The thin-section MRIs of all patients were interpreted in consensus by 2 neuroradiologists (D.K. and Y.J.C., with 5 and 13 years of clinical experience in neuroradiology, respectively). Before evaluation, the 2 neuroradiologists completed a training session on imaging from 5 patients to help them reach a consensus on evaluation of the imaging findings. The 2 neuroradiologists were blinded to other radiologic imaging including the DSA, and to clinical information including symptoms, signs, and treatment given to the patients. The following possible imaging predictors for CCF on pretreatment thin-section MR imaging were analyzed: 1) abnormal contour of the cavernous sinus, 2) abnormal internal signal void of the cavernous sinus, 3) prominent venous drainage flow (anterior/lateral/posterior), and 4) orbital/periorbital soft tissue swelling.
Abnormal contour of the cavernous sinus was considered positive if the cavernous sinus was asymmetrically enlarged and the lateral wall of the cavernous sinus was outwardly convex or straight on the coronal view (Fig 2). Abnormal internal signal void of the cavernous sinus was considered positive if there was a signal void larger than 2 mm in the shortest diameter within or along the wall of the cavernous sinus on coronal imaging (Fig 3). Signal void was evaluated on both T1-and T2-weighted MR imaging. Prominent venous drainage flow was considered positive if the superior or inferior ophthalmic vein (anterior), sphenoparietal sinus (lateral), or superior or inferior petrosal sinus (posterior) was prominently or asymmetrically enlarged with or without signal void (Figs 4 and 5). Orbital/periorbital soft tissue swelling was considered positive if there was high signal change and extraocular muscle thickening on T2-weighted MR imaging (Fig 6). The overall diagnosis (presence or absence of CCF) was also evaluated for each case, and the diagnostic performance of thin-section MR imaging was estimated.
The DSAs of all patients were reviewed by a neurointervention-dedicated neuroradiologist (Y.S, with 5 years of clinical experience in neuroradiology). The neuroradiologist was blinded to other radiologic images, including the thin-section MR imaging, and to clinical information, including symptoms and signs of the patient and treatment performed. The presence or absence of CCF on DSA was evaluated and considered as the reference standard. When present, the type of CCF (whether direct or indirect) was also evaluated. For patients with indirect CCF, a further classification according to Suh et al 33

Statistical Analysis
This study had 2 main outcomes. The primary outcome was the diagnostic performance of thin-section MR imaging for the diagnosis of CCF with the reference standard being DSA. The sensitivity, specificity, accuracy, positive and negative predictive values, and appropriate use criteria of the receiver operating characteristic curves of thin-section MR imaging for depicting the Coronal T2-weighted image of a patient with diplopia, confirmed to be left sixth cranial nerve palsy on neurologic examination. Note the abnormal contour bulging of the left cavernous sinus (arrow). An internal signal void was also noted on both T2-weighted (arrow) and T1-weighted imaging (not shown). The patient was confirmed as having a direct CCF on digital subtraction angiography.  presence of CCF were calculated. The secondary outcome was to determine possible imaging predictors for CCF on thin-section MR imaging. The Fisher exact test and univariable logistic regression analysis were used to determine possible predictors of CCF. The thin-section MR imaging characteristics of abnormal contour of the cavernous sinus, internal signal void of the cavernous sinus, prominent venous drainage flow (anterior/lateral/ posterior), and orbital/periorbital soft tissue swelling were considered as potential adjustment variables. Statistical analyses were performed by using SPSS software (version 25.0; IBM), with statistical significance being defined as P , .05.

RESULTS
Among the 98 patients, 38 were confirmed as having CCF on DSA. Six patients had direct CCF, and 32 patients had indirect CCF. The baseline characteristics and clinical manifestations of the patients are summarized in Table 1. The overall diagnostic performance of thin-section MR imaging for diagnosing CCF was as follows: accuracy, sensitivity, and specificity were 88.8%, 97.4%, and 83.3%, respectively. There was 1 false-negative case (3.1%, 1/38) on thin-section MR imaging. Among the 32 indirect CCF cases, there were 10 proliferative type cases, 17 restrictive type cases, and 5 late restrictive type cases, while the single false-negative case was a late restrictive type. Among the 60 patients confirmed as CCF-negative on DSA, there were 10 false-positive cases on thin-section MR imaging (16.7%, 10/60).
In the univariable logistic regression, patients with CCF showed higher odds of an abnormal contour of the cavernous  Table 2.
Radiologists do not usually rely on a single imaging feature to diagnose a disease; therefore, we analyzed various combinations of features to determine the best ones for diagnosis. Among the various feature combinations showing an accuracy above 85%, "any prominent venous drainage flow" and "internal signal void of the cavernous sinus" (combination 1) showed the highest accuracy (91.8%); "any prominent venous drainage flow" (anterior/ lateral/posterior) and "orbital/periorbital soft tissue swelling" (combination 2) showed the highest specificity (96.7%) and positive likelihood ratio (23.7); and "prominent anterior venous drainage flow" or "orbital/periorbital soft tissue swelling" (combination 3) showed the highest sensitivity (92.1%) and the lowest negative likelihood ratio (0.097). The diagnostic performances of each feature and combination of features are summarized in Table 3.

DISCUSSION
In this study, we analyzed the diagnostic performance of thinsection MR imaging for CCF, and determined possible thinsection MR imaging predictors for CCF, by using a historical cohort of patients with clinically suspected CCF. The accuracy, sensitivity, and specificity of thin-section MR imaging for diagnosis of CCF were 88.8%, 97.4%, and 83.3%,  Most of the research focusing on the MR imaging findings of CCF was conducted between the late 1980s and early 1990s. [9][10][11][12][13] The MR imaging findings of these studies included low or dark signal within the cavernous sinus, dilated superior ophthalmic vein, swelling of the extraocular muscles, and bulging of the lateral wall of the cavernous sinus. Uchino et al 12 analyzed 10 cases of indirect CCF and found that 9 out of 10 cases (90%) were positive for flow void within the cavernous sinus, and 8 out of 10 cases (80%) were positive for dilated superior ophthalmic vein on MR imaging. These findings correlate well with the results of our study, in which 92.1% of the CCF cases showed internal signal void within the cavernous sinus, and 78.9% showed prominent anterior venous drainage flow. However, most previous research covers case series studies in which the number of cases was less than or equal to 10. The study protocols used in previous studies mainly included non-3D T1-weighted, T2-weighted, and postcontrast T1-weighted imaging with section thickness ranging from 2 to 10 mm, acquired on scanners with magnetic fields equal to or less than 1.5T. To our knowledge, there has been no study on the diagnostic performance of MR imaging for CCF that has included a large number of cases. There was 1 false-negative case on thin-section MR imaging, which was an indirect CCF (late restrictive type). The late restrictive type CCF is characterized by a few arterial feeders and sluggish retrograde venous flow on DSA, 33 and these hemodynamic characteristics may be the cause for the incorrect diagnosis on thin-section MR imaging. There were 10 false-positive cases (16.7%) in terms of overall diagnostic performance in our study. In the false-positives, the common image findings were signal void in the cavernous sinus (10 cases), abnormal contour of cavernous sinus (9 cases), and prominent venous drainage flow in the lateral (4 cases) and posterior (4 cases) venous drainage structures. In most of our study cases, where fistulous or venous flow was not very slow, thin-section MR imaging showed satisfactory diagnostic performance in patients with clinically suspected CCF, with especially high sensitivity. Our study also focused on evaluating the possible imaging predictors for CCF on thin-section MR imaging. Abnormal contour of the cavernous sinus (OR: 21.7), internal signal void of the cavernous sinus (OR: 15.3), prominent venous drainage flow in anterior, lateral, or posterior drainage venous structures (OR: 54.0), and orbital/periorbital soft tissue swelling (OR: 40.4) on thin-section MR imaging were shown to be possible imaging predictors for CCF on thin-section MR imaging. Abnormal contour and signal void of the cavernous sinus were visible on coronal T1WI and T2WI with 3-mm thickness. Prominent venous drainage flows were best seen on the contrast-enhanced 3D T1WI with 0.6 mm thickness and axial reconstruction. Orbital/periorbital soft tissue swelling was well visualized on coronal images with 3-mm thickness, whereas it was not clearly visualized on contrast-enhanced 3D T1WI. It is evident that thin-section MR imaging is helpful in searching for these fine struc-  tures and lesions, though determination of the optimal section thickness may require additional study.
There are several limitations to this study. First, the study was retrospective and patients were selected on the basis of their electronic medical records and the presence of certain radiologic examinations. This might have resulted in a selection bias. Furthermore, the MR imaging reviewers were aware that the patients had undergone DSA, and this might have induced some reader bias and might have affected diagnostic performance. Second, because this was a single-center study, the generalizability of these results may be limited, and a further prospective multicenter study may be needed. Third, there was a delay of up to 90 days between the MR imaging and DSA. However, in most cases the delay was less than 30 days (86.7%), and considering the slow-evolving natural course of CCF, this delay may be acceptable. Fourth, the comparison of thin-section MR imaging with CTA or MRA was not conducted. CTA and 3D TOF-MRA are also known as good modalities for screening CCF with high sensitivities. However, the case numbers of concurrent CTA with thin-section MR imaging and 3D TOF-MRA with thin-section MR imaging were too small to perform the comparative analysis in our study. Therefore, the results of this study do not justify the replacement of CTA or MRA to thin-section MR imaging in case of clinically suspected CCF. However, the results may be helpful in differentiating the cause of orbital or cranial nerve symptoms in patients who have undergone the thin-section MR imaging without CTA or MRA.

CONCLUSIONS
Thin-section MR imaging showed high performance for diagnosing CCF in patients with clinically suspected CCF. The possible imaging predictors for CCF on thin-section MR imaging included abnormal contour of the internal signal void in the cavernous sinus; prominent venous flow in the ophthalmic vein, sphenoparietal sinus, or petrosal sinus; and orbital/periorbital soft tissue swelling.