Article Text
Abstract
Background To date, very little study of the importance of a volumetric T2-weighted MR sequence in the evaluation of spinal vascular malformations (SVMs) has been carried out.
Objective To determine the utility and accuracy of a volumetric T2 MR sequence compared with conventional T2 in the diagnosis of SVMs.
Methods Retrospective analysis of all patients who underwent spinal DSA for suspected SVMs was conducted. Conventional T2 and volumetric T2 MR images were analysed for the presence of flow voids and parenchymal changes, and SVMs were characterized. The sensitivity, specificity, and overall diagnostic accuracy of these MRI diagnoses were calculated.
Results Of 89 subjects included in the final analysis, 70 patients had angiographically proved SVMs (38 patients with spinal cord arteriovenous malformations [SCAVM—intramedullary or perimedullary] and 32 cases of spinal dural arteriovenous fistula (SDAVF)) and the remaining 19 subjects were normal. The sensitivity and specificity for identification of SVMs were 98.1% and 90% for volumetric T2 sequences, compared with 82.8% and 89.4% for conventional T2 MRI, respectively. For characterization of spinal vascular lesions, volumetric MRI showed high sensitivity, specificity, and accuracy for SDAVF (100%, 90%, 97%, respectively) compared with conventional T2 MRI (71.8%, 89%, 79%, respectively). The positive likelihood ratio was high and negative likelihood ratio was zero for volumetric MRI evaluation of SDAVF, while these ratios were comparable between the two sequences for SCAVM.
Conclusion Volumetric T2 MRI is highly sensitive for the detection of SVMs, especially for SDAVF. Volumetric T2 MRI could be introduced into routine clinical practice in the screening of suspected SVMs.
- arteriovenous malformation
- magnetic resonance angiography
- spine
Statistics from Altmetric.com
Spinal vascular malformations (SVMs) comprise of a rare group of spinal vascular lesions such as spinal dural arteriovenous fistula (SDAVF), intramedullary arteriovenous malformations (AVMs) and perimedullary arteriovenous fistula, and account for about 10% of all central nervous system vascular malformations.1 Spinal MRI plays an important role as an initial screening imaging modality in patients with suspected vascular myelopathies to confirm or rule out SVMs, based on which a decision to obtain a spinal angiogram is usually made. Spinal angiography (SA) is the ’gold standard' imaging technique for the diagnosis of SVMs; however, it is an invasive, time-consuming technique that involves exposure to high levels of radiation and requires a large volume of contrast agent. The radiation dose for SA may vary between 6.5 and 13 mSv. This is about three times the radiation exposure with diagnostic cerebral angiography and the contrast volume injected may range between 16 and 245 mL.2 3 Although the technique is associated with a low incidence of neurologic and systemic complications, serious adverse events (<1%) have been reported.2 Atherosclerosis or degenerative aorta may increase the risk of complications or the need for repeated examinations.2 4
A conventional T2-weighted sequence is the initial screening tool used to triage patients for further evaluation with a spinal angiogram. Large shunts or high flow vascular malformations such as AVMs can be easily identified by the presence of characteristic flow voids. However, the slow flow vascular malformations such as SDAVF or spinal perimedullary fistulas can often pose a diagnostic challenge owing to their ambiguous or subtle imaging characteristics. When MRI is negative but the clinical suspicion is strong, SA is still performed to avoid missing a potentially treatable cause of myelopathy. However, SA is not infallible and the diagnosis of SVMs, especially SDAVF can still be missed, if strict adherence to the technical standards for performing and interpreting angiograms is not followed.5 Inadequate injection into the feeder, non-inclusion of the region of interest, omission of segmental arteries, or overlooked dorsospinal arteries are identified as potential factors contributing to a false-negative SA result.5 6 Improving the accuracy of imaging could thus greatly aid in the diagnosis of SDAVF, as it improves the diagnostic confidence of the angiographer in ascertaining the diagnosis.
The diagnostic performance of conventional T2 MR imaging and time-resolved, contrast-enhanced MRI in the evaluation of SVMs has been reported previously.7–11 However, the role of a heavily T2-weighted volumetric MR sequence in the initial evaluation of SVMs has been less well studied. Theoretically, owing to its superior spatial resolution, reformatability, and pan-vertebral coverage, a high-resolution sequence is expected to detect subtle flow voids along the entire spinal axis, which would otherwise be missed in the routine MR sequences.12 Thus, it is expected to improve the overall diagnostic accuracy of SVM detection. In this report, we evaluated the usefulness of both a conventional and a volumetric T2 sequence in the evaluation of SVMs.
Materials and methods
A retrospective case–control study was conducted. All patients with suspected SVMs, evaluated at our institution during a 10-year period from January 2007, were included in the analysis. The suspicion was based on clinical and radiological observations. These patients were initially investigated with spinal MRI in a 1.5 T or 3 T MRI machine, which was followed by SA. Inclusion criteria were (1) a spinal angiogram recording the presence or absence of SVMs; (2) availability of conventional and/or high- resolution, T2-weighted imaging that included the entire spinal neuroaxis and spinal canal; (3) imaging study performed within 2 weeks of SA. Subjects were excluded if the quality of the images was suboptimal. The study was approved by institutional ethics committee, with patient consent waived as the study was retrospective.
The MRI protocol included a 3D volumetric turbo spin-echo, T2-weighted sequence(supplementary file), either 3D T2 SPACE (Sampling Perfection with Application Optimized Contrasts using different flip-angle Evolutions), obtained in 1.5 T Siemens SQ Avanto, or 3D T2 CUBE carried out in a 3 T GE Discovery 750E. Parameters of 3D volumetric T2 sequences were as follows: T2 SPACE: TR/TE, 1500 ms/143 ms; NEX, 2; FOV, 280 mm; flip angle, 150°; slab thickness, 0.9 mm; matrix, 370×320; and total time of acquisition, 5.26 min. T2 CUBE: TR/TE, 2500 ms/113 ms; NEX, 2; FOV, 300 mm; flip angle, 150°; slab thickness, 1.6mm; matrix, 288×288; and total time of acquisition 6.09 min. Post-contrast imaging was not routinely performed. MR images were acquired at two or three levels to cover the entire spine and these images were reconstructed by fusion software provided by the MR vendor. 3D volumetric MR images were further analysed in multiple planes in a vendor or PACS (picture archiving and communications system) workstation. The interval between MRI and SA was limited to <2 weeks in all patients. Both conventional and volumetric studies were available for review in 73% of included patients.
Subjects with angiographically proved vascular malformations were categorized as cases and the remaining subjects as controls. Conventional T2 followed by volumetric T2 image datasets were analysed separately and consensus reached by two neuroradiologists with 9 and 2 years' experience, respectively, who were blinded to the final diagnosis. Images were anonymized and read with a time gap of 2 weeks to exclude memory bias. MRI scans were evaluated for the presence of vascular flow voids (intramedullary or perimedullary), cord hyperintensity, cord expansion, and presence of nidus, based on which the SVMs were identified and classified. A spinal angiogram was independently analysed by a neurointerventional radiologist of 14 years’ experience and findings were compared with the MR diagnosis.
Statistical analysis
Sensitivity, specificity, true positivity, true negativity, false positivity, and false negativity were analysed separately for volumetric and conventional T2 MR, with SA being the gold standard. Positive and negative likelihood ratios for the presence of flow voids in the identification of SVMs were also calculated. Interobserver agreement for various imaging indicators was calculated using κ statistics. A p value <0.05 was considered as statistically significant. All analyses were performed with SPSS software, version 22.
Results
A total of 89 subjects who fulfilled the inclusion criteria were included in the final analysis (figure 1 and table 1). Angiographically proved SVMs were found in 70 patients and the remaining 19 patients had no evidence of SVMs(supplementary file). Thirty-two (45.7%) of the 70 subjects had SDAVF, which included 27 men (84%) and five women (16%), with an average age of 55.4 years (range 27–76). In 38/70 (54.3%) patients, SCAVM was diagnosed, comprising 26 (68%) cases of intramedullary AVMs and 12 (32%) cases of perimedullary fistula. The mean age of patients with SCAVM was 35.6 years (range 13–66), with a male to female ratio of 3.2. There were no complications associated with SA.
The sensitivity and specificity of intramedullary or perimedullary flow voids (table 2) in the identification of SVM were 98.1% and 90%, respectively, for volumetric T2 sequence, compared with 82.8% and 89.4%, respectively, for conventional MRI. For detection of flow voids in SDAVF, volumetric T2 had 100% sensitivity while conventional T2 had only 71.8% sensitivity. For SCAVM, the sensitivity of volumetric T2 (96.4%) was better than that for conventional T2 (92%). There was no significant difference in the nidus detection rate of intramedullary AVMs, between conventional (77%) and volumetric T2 MRI (78.9%). Overall, volumetric T2 was more accurate than conventional the T2-weighted sequence for both SDAVF (0.97 vs 0.79) and SCAVM (0.94 vs 0.91), though the effect was more pronounced for SDAVF(figure 2 and 3). The positive likelihood ratio (table 3) for SDAVF or SCAVM in the presence of flow voids varied between 9 and 10, suggesting a high probability of a positive diagnosis with SA. The negative likelihood ratio was 0 for SDAVF; however, owing to one false-negative diagnosis in a patient with low-flow spinal perimedullary fistula, the negative likelihood ratio was 0.04 for SCAVM. A good agreement was noted between the two readers for detection of flow voids in both the imaging modalities, though it was higher for volumetric T2 MRI (κ value: 0.81 vs 0.77). The independent assessment of radiologists for detection and characterization of SVMs is shown in table 4.
Other ancillary findings evaluated included cord expansion, cord hyperintensity, and parenchymal hemorrhage. Among 19 subjects with normal spinal angiograms, 11 (57.9%) showed cord hyperintensity and 5 (26.3%) showed cord expansion. Of the 32 patients with SDAVF, 27 (84.4%) had cord expansion and 29 (90.6%) had cord hyperintensity. Eight of the 12 (66.7%) subjects with perimedullary fistula diagnosed by DSA, showed cord expansion and hyperintensity. Of the 26 patients with spinal AVMs, 23 (88.5%) showed focal cord expansion, whereas only 19 (73.1%) patients with AVMs had cord hyperintensity. No statistically significant difference was noted between the two imaging sequences (p<0.05).
Discussion
Our study showed that the T2 high-resolution volumetric sequence had very high sensitivity and accuracy in the evaluation of SVMs as a whole and achieved 100% sensitivity in the detection of SDAVF. A high positive likelihood ratio and very low negative likelihood ratio for the presence of flow voids for diagnosis of SVMs indicated that this sequence could be employed reliably, as an initial investigational tool for suspected SVMs.
Several imaging strategies are adopted in the evaluation of SVMs, including conventional MRI, contrast- enhanced MRI, time-resolved contrast MRI sequences such as TRICKS (Time-Resolved Imaging of Contrast Kinetics), and high-resolution T2 sequences. Initially, patients are often evaluated with conventional T2 MRI sequences. The two important observations that help in the diagnosis of spinal arteriovenous fistula are the presence of flow voids and cord hyperintensity. In a study by Toossi et al,13 the presence of these two findings had 100% sensitivity in diagnosing SDAVF; however, both signs were noted in approximately 70% of the patients only. Flow voids, which are characteristic of SVMs, were observed in 81% of the population. Recommendation for SA based on these observations varied between 86% and 94%, indicating that information gleaned from imaging may be inadequate to triage all patients.13 The sensitivity of conventional MRI is especially low for slow-flow vascular malformations such as SDAVF or perimedullary fistula compared with high-flow vascular malformations such as cord AVMs. In a large study involving 115 patients, high-flow AVMs were correctly identified in 96.6% of patients, while accurate diagnosis of slow-flow vascular lesions was possible in only 38.4%.7 The importance of identifying flow voids cannot be overestimated as in the absence of this finding, 40% of the slow-flow lesions were misdiagnosed as other spinal pathologies. Conventional MRI is prone to miss these subtle flow voids owing to thick imaging slices, interleaving gap, or inadequate coronal coverage.
Time-resolved MR angiographic techniques have been used in the evaluation of SVMs.9–11 In this technique, the centre of K space is acquired more frequently than the periphery so that rapid image acquisition with high temporal resolution of arterial and venous phases can be obtained. Typically, with this technique, more than 20 angiographic phases of images can be acquired at a rate of 1–2 frames/s. In a study by Amarouche et al, it was observed that time-resolved contrast MR angiography techniques had 98% sensitivity and 63% specificity in the evaluation of SVMs.10 Although SDAVF could be correctly recognized and categorized using this technique, false-positive and false-negative results were noted in the diagnosis of SCAVM. Other studies evaluating SDAVF with time-resolved MR angiography also demonstrated sensitivity and specificity of 88–100% and 88–90%, respectively.11 14 Although it is an excellent modality for detecting, characterizing, and planning intervention of SVMs, its restricted length of coverage could be a serious disadvantage, especially in lesions located outside the field of view. The slow-flow SVM with venous drainage away from the field of view could be missed altogether, leading to a false-negative result.
Another MR technique employed is volumetric T2-weighted MRI (SPACE, CUBE or VISTA). Although this sequence is available from all MRI vendors, only a few studies have discussed its usefulness in the evaluation of SVMs. This sequence is used for both diagnosis and preprocedural planning of SDAVF treatment.12 15 16 In patients with confirmed SDAVF, this sequence was found to have 94% accuracy in localizing the feeder to one vertebral level. Comparing conventional MRI and volumetric T2 MRI in assessing SDAVF, Kralik et al 15 showed that both sequences achieved 100% sensitivity for diagnosis, though false-positive observation was a concern in the latter group. However, this study had a few limitations, such as a relatively small sample size in the 3D cohort, lack of direct comparative analysis between 2D and 3D groups, and non-inclusion of other vascular malformations, such as perimedullary fistula or cord AVMs. Thus, the significance of the effect of the 3D sequence is unclear, in addition to the 2D sequence, in SDAVF evaluation or whether these observations could also be generalized to other SVMs.
For several reasons a volumetric T2-weighted sequence is specifically suited to identifying vascular flow voids. Owing to high spatial resolution and isotropic data acquisition, multiplanar reconstruction in any plane is feasible and the suspected flow voids and their feeders could be evaluated in detail before arriving at a conclusive diagnosis. Also, owing to excellent contrast between the cerebrospinal fluid and cord parenchyma, even subtle flow voids in the subarachnoid space could be identified and analysed for authenticity. In our study, all cases of SDAVF were correctly identified on a volumetric T2-weighted sequence. There was no false-negative diagnosis in the SDAVF cohort, but one such diagnosis was made in the SCAVM cohort, where a slow-flow type 1 perimedullary fistula was not detected owing to the absence of flow voids. The results are particularly encouraging in comparison with previous studies reporting the efficacy of conventional MRI in the diagnosis of SVMs.4 10 The results are also comparable to those obtained with other advanced imaging modalities, such as time-resolved MR angiography.10 11 14 Our study demonstrated that the presence of flow voids had high likelihood ratio of 9 to 10 for SDAVF or SCAVM and hence, the presence of such imaging findings should elicit a high clinical suspicion and thorough search for occult SVMs, before ruling out the SVM diagnosis. A non-zero negative likelihood ratio points to the limitation of this particular sequence in the detection of flow voids in very slow-flow SVMs and hence, in an appropriate clinical setting, further imaging and spinal angiography is still warranted.
While interpreting images of suspected myelopathy, physicians can err towards a false-positive diagnosis in order to avoid missing a potentially treatable cause of myelopathy. Indeed, in our study, a false- positive rate of 10% was observed. However, we believe that it can be minimized if thorough evaluation of flow voids with respect to their feeders, nidus, or draining vein is performed. This will also aid in arriving at definitive pathological diagnosis and enable planning a further course of management.16 A good agreement between the two observers of varying experience was noted in our study and the accuracy of diagnosis between them was also comparable (0.94 vs 0.97). This observation suggests that there was no learning curve in interpretation, and general radiologists in non-referral centres could also make a diagnosis with reasonable confidence if suspicion of SVMs was high. One of the purported disadvantages of the volumetric T2 sequence is the inherent poor tissue contrast, so that signal abnormalities of cord can become less conspicuous.9 12 Since parenchymal signal alteration is a useful sign in the diagnosis of SVMs, diagnosis relying solely on the presence of flow voids could lead to false interpretations, especially when clinical history is not indicative of the diagnosis or when flow voids are highly subtle. A previous study noted that cord changes were observed in 67% with volumetric T2 MRI as opposed to 83% in conventional MRI.15 In our study, cord signal changes could also be assessed in the volumetric sequence and there was no significant difference from the conventional sequence, indicating that parenchymal evaluation is not a real limitation of this sequence.
Our study has several strengths. The study population is large and comprised different spinal vascular pathologies such as SDAVF, perimedullary arteriovenous fistula, and SCAVM. The images were interpreted by two blinded readers with widely varying experience, which permitted study of the impact of a learning curve on diagnostic performance. Also, the imaging protocol remained uniform throughout the study.
Some of the limitations of our study include its retrospective study design and the influence of referral bias in image interpretation. Being a major referral centre for SVMs, readers are aware of the high likelihood of a positive diagnosis and are likely to perform a thorough search for diagnosis. Thus it remains to be seen in further studies, whether the high sensitivity achieved in our study could also be reproduced in peripheral low-volume centres. In this study, the presence of flow voids was found to be highly accurate in distinguishing vascular from non-vascular pathologies, thus the potentially additive role of imaging features such as parenchymal changes and cord expansion on the diagnosis were not evaluated. However, whether this high discriminatory power is due to the smaller sample size of the control group or reflects diagnostic superiority of high-resolution sequences needs further clarification. Also, a one to one comparison of conventional and volumetric sequences was not performed. Thus, though 3D sequences could be considered as a one-stop solution for suspected SVMs, their role as a screening test is uncertain until further studies demonstrate the real incidence or imaging characteristics of flow voids in a large sample size of the normal population or patients with unrelated pathologies. If this known, unnecessary invasive angiography could be avoided in such populations with positive findings and extremely low clinical suspicion.
Conclusion
Volumetric T2-weighted MRI is highly sensitive and accurate for the detection and characterization of SVMs. Inclusion of a volumetric T2-weighted MR sequence in routine imaging practice could minimize misdiagnosis in patients with suspected SVMs.
Supplemental material
Acknowledgments
Part of this study was presented at the 103th RSNA conference at Chicago, USA.
References
Footnotes
Contributors Concept and study design: SKK; acquisiton of data: SKK, AR; analysis of results and interpretation: all authors; manuscript preparation: SKK, BT; manuscript revision and final approval: all authors.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement The data can be shared upon request by email to corresponding author.