Introduction

Cerebral arteriovenous malformations (CAVMs) are congenital vascular abnormalities that connect arteries with veins, and they can cause fatal cerebral and subarachnoid hemorrhages [1, 2]. However, surgical treatment of CAVMs sometimes leads to serious complications, so clinicians tend to carefully consider the indications for surgery. A recently published randomized trial of unruptured brain arteriovenous malformations (the ARUBA study) just increased this tendency [3]. To help assess the risk of hemorrhage and to determine treatment strategies, imaging studies involving the detailed structures of CAVMs are required. The following factors have been reported as risk factors for cerebral hemorrhage due to CAVM surgery: history of hemorrhage, localization of nidus, aneurysm, location in a deep part of the derived vein, and high Spetzler–Martin grade [4,5,6,7,8,9].

To evaluate CAVMs, digital subtraction angiography (DSA) is the gold standard radiological technique. However, this method is not suitable for screening and follow-up of CAVMs in routine medical care because it can cause thromboembolic complications and involves contrast media and radiation exposure [10,11,12]. Therefore, 3D time of flight (TOF)-magnetic resonance angiography (MRA) is often used instead of DSA to follow up patients with CAVMs because of its less invasiveness. However, the method confers insufficient visualization of certain vascular components of CAVMs [13, 14], and cannot detect especially small lesions, which, for example, frequently occur in hereditary hemorrhagic telangiectasia (HHT) [15, 16].

MRA using the silent MRA algorithm (silent MRA), which combines arterial spin labeling (ASL) and an ultrashort time echo (UTE), was developed in 2012 [17]. Although ASL-based MRA has been reported to be useful in the evaluation of CAVMs [18,19,20,21,22,23,24,25,26], many studies were about ASL-based four-dimensional (4D) MRA techniques using different inversion time [18,19,20,21]. Silent MRA could only visualize a single time point which is very simple and approachable for clinicians. In addition, UTE techniques may have advantages over ASL-based 4D MRA because it minimizes dissipation of the labeled intracranial blood flow signal and alleviates magnetic susceptibility artifacts. To our knowledge, however, the usefulness of silent MRA in the evaluation of CAVMs has not been investigated. Therefore, this study aimed to determine whether silent MRA is useful for evaluating CAVMs.

Materials and methods

Study population

The study retrospectively recruited consecutive patients with CAVMs; The inclusion criteria were as follows: patients of all ages who were diagnosed by 4D CT angiography (CTA) or DSA, and who underwent the two types of MRA at Keio University hospital between August 2015 and August 2018. Of 37 CAVM patients who met the inclusion criteria, we excluded 4 patients who had a complete cure of AVMs on DSA after endovascular treatment and 5 patients whose TOF-MRA images did not cover whole parts of AVM. One patient had double lesions. Thus, 29 CAVMs of 28 patients were included in this study. The mean age of the patients was 47 years (range 15–78 years). The study population mostly included males (16/28). The mean Spetzler–Martin grade was 2.24 ± 0.95 (Grade 1, 10; Grade 2, 9; Grade 3, 8; Grade 4, 2; Grade5, 0). The mean size of CAVMs was 23.8 ± 19.1 mm (1.5–67 mm). Ten micro AVMs with a nidus diameter less than 10 mm were included, and 66% of patients were incidentally diagnosed (other presentations include: 5 cases, hemorrhage; 4 cases, epilepsy). Written informed consent was obtained from the patients or their families.

MRA technique

Three-dimensional TOF-MRA and silent MRA were performed on the same test day using a 3-T magnetic resonance imaging (MRI) system (SIGNA Pioneer; GE Healthcare, Milwaukee, Wisconsin) and a 24-channel head coil. The scan parameters of TOF-MRA were as follows: matrix, 384 × 329; field of view (FOV), 18 × 18 cm; number of slabs, 4; number of overlapped sections between two slabs, 11; flip angle, 18°; TR, 23 ms; TE, 2.9 ms; section thickness, 1.0 mm; number of excitations (NEX), 1; bandwidth, 31.25 kHz; and acquisition time, 6 min, 14 s. Those of silent MRA were as follows: matrix, 150 × 150; FOV, 18 × 18 cm; flip angle, 5°; TR, 799 ms; TE, 0.016 ms; section thickness, 1.2 mm; NEX, 1; bandwidth, 31.25 kHz; and acquisition time, 5 min, 29 s. The readout-sampling scheme in silent MRA was zero TE sequence by 3D radial sampling. The label method and duration have not been disclosed to the public, and we used silent MRA with default settings used in the previous studies [27]. The slice geometry of TOF-MRA and silent MRA were OM-line and axial plane, respectively. All MRA scans analyzed in the present study were performed within 2 months before or after DSA or 4D CTA.

DSA, 4D CTA technique

After catheterization of the internal carotid and the vertebral arteries via a femoral artery approach, diagnostic biplanar intra-arterial DSA (Inova; GE Healthcare, Milwaukee, Wisconsin, USA) was performed using automated injection of a contrast agent by a trained neurosurgeon. Temporal resolution of the images was 4 frames/s and 7.5 frames/s if necessary. 4D CTA was performed using a 320-slice scanner (Aquilion One; Canon Medical Systems, Otawara, Japan). To determine the optimal timing of dynamic scans, a test scan was performed using 20 mL contrast agent of 370 mg I/mL (iopamidol, Iopamiron 370; Bayer Yakihin, Osaka, Japan) at 5 mL/s, followed by 20 mL of saline, at the level of the carotid bulb. The 4D CTA scanning protocol consisted of a 11-s continuous acquisition (80 kV, 300 mAs, 0.5-s rotation speed) in the arterial to venous phase after a bolus injection of 50-mL contrast agent and 4-volume intermittent scans at a 4-s interval in the late venous phase.

Image analysis

One neuroradiologist (22 years of experience) selected the cases which met the inclusion criteria for the reading study. Confirmation by means of DSA or 4D CTA was used as the reference standard for all the evaluations. With regard to the AVM detection and visualization, two experienced interventional neurosurgeons (both had 7-years experiences of DSA and MRI) independently reviewed TOF-MRA and silent MRA images on a PACS workstation. Disagreements were resolved by consensus. Three-dimensional maximum intensity projection (MIP) and source images were used for the interpretation of the MRA images. The order of images to be interpreted was TOF-MRA first, followed by silent MRA. The interval of readings between TOF-MRA and silent MRA was over at least 2 weeks. The observers were blinded to other imaging findings and clinical information. The criteria for a detection of AV shunt lesions on TOF-MRA and silent MRA images included the presence of hyperintense signal in areas of the nidus and/or drainer. Micro AVM was defined as the AVM with a nidus diameter less than 10 mm.

The observers independently graded the visualization of the feeder, nidus, and drainer of AVMs on TOF- and silent MRA images using a 5-point scale: Grade 1, not visible (no signal); Grade 2, poor (ambiguous visualization with severe blurring or artifacts); Grade 3, fair (moderate image quality with moderate blurring or artifacts); Grade 4, good (good image quality with slight blurring or artifacts); and Grade 5, excellent (sufficient image quality without artifacts).

Assessment of Spetzler–Martin grade by using TOF-MRA and silent MRA was independently performed by the two observers. Disagreements were resolved by consensus. CAVMs were characterized according to the Spetzler–Martin grading system [28]. The nidus size was classified as small (< 3 cm), medium (3–6 cm) or large (> 6 cm). Measurements were performed on source images of MRA. Location was categorized as being in an eloquent or non-eloquent area and venous drainage was categorized as superficial or deep. When both superficial and deep venous drainages were seen, it was classified as deep. In the assessment of Spetzler–Martin grade, T2-weighted MRI was also used as the standard reference.

After the blinded study, the structures of CAVMs on MRA images were reviewed retrospectively by the two observers in consensus, together with the clinical and angiographic findings.

Statistical analysis

The detectability of AV shunt lesions on TOF- and silent MRA images was calculated using DSA or 4D CTA as reference standards. The accuracy of Spetzler–Martin grading was compared between TOF-MRA and silent MRA. Fisher’s exact test was performed to compare the accuracy between the results of TOF-MRA and those of silent MRA. To compare the mean visualization scores between the TOF-MRA and silent MRA, the Mann–Whitney U test was used. The level of interobserver agreement between the two readers with respect to the visualization of the AVM was determined by calculating the k coefficient (k < 0.2, poor; 0.21 < k < 0.40, fair; 0.41 < k < 0.60, moderate; 0.61 < k < 0.80, good; 0.81 < k < 0.90, very good; and k > 0.90, excellent agreement).The software R was used for all analyses [29], A p value of less than 0.05 was considered to indicate a statistically significant difference.

Results

For all 29 CAVMs, 23 (79%) lesions were detected for TOF-MRA (k = 1.0) and all (100%) for silent MRA (k = 1.0). Of 10 micro AVMs, only 4 (40%) lesions were detectable on TOF-MRA (k = 1.0) and all the lesions on silent MRA (k = 1.0). A significant difference in the detection rate of micro AVMs between the sequences was observed (p < 0.01).

Table 1 and Fig. 1 show the summary of the visualization score of the feeder, nidus, and drainer on TOF-MRA and silent MRA images. The mean score ± standard deviation of feeder, nidus, and drainer was 3.21 ± 1.44, 2.07 ± 164 0.84, and 1.86 ± 1.06 for TOF MRA and 3.66 ± 1.32, 4.24 ± 0.72, and 3.17 ± 1.47 for silent MRA, respectively. The mean score of the nidus and drainer was significantly higher for silent MRA than TOF-MRA (p < 0.001), while there was no significant difference in the feeder between the two sequences (p = 0.22) (Fig. 2). The interobserver agreement was moderate-to-good for both TOF-MRA and silent MRA (Table 1). The similar results were also found when assessing micro CAVMs (Table 2 and Fig. 3). The summary of the assessment of Spetzler–Martin grade by TOF-MRA and silent MRA is shown in Table 3. The accuracy of Spetzler–Martin grade for TOF-MRA and silent MRA were 11/29 (38%) and 23/29 (79.3%), respectively. There was a significant difference in the accuracy between the two types of MRA (p < 0.001).

Table 1 Visibility of cerebral AVM components on two types of MRA images
Full size table
Fig. 1
figure 1

Distribution of the image quality grade for feeder (a), nidus (b) and drainer (c) of AVM on TOF-MRA and silent MRA images. For feeders on both the MRA sequences, 80–90% were delineated with acceptable quality (Grades 3–5). The delineation with acceptable quality was obtained in about 95% of all niduses for Silent MRA, whereas it was 30% for TOF MRA. For drainers, the delineation with acceptable quality was seen in more than 60% of lesions for silent MRA and in only 30% for TOF MRA. Grade 1, not visible (no signal); Grade 2, poor (ambiguous visualization with severe blurring or artifacts); Grade 3, fair (moderate image quality with moderate blurring or artifacts); Grade 4, good (good image quality with slight blurring or artifacts); and Grade 5, excellent (sufficient image quality without artifacts)

Full size image
Fig. 2
figure 2

DSA and MRA images of a 40-year-old woman with a left occipital CAVM. DSA (lateral projection from the left internal carotid artery) at arterial (a) and late arterial phases (b) reveals a AVM nidus (*) fed primarily by the left posterior cerebral artery (arrowhead) and its drainer (arrow). The nidus (asterisk) and drainer (arrow) are better delineated on Silent MRA images (d) than the TOF MRA ones (c) (arrows). The feeders were judged to be Grade 5 on TOF- (c) and Silent MRA images (d) by both observers. The drainers were judged to be Grade 2 on TOF- (c) and Grade 5 on silent MRA images (d)

Full size image
Table 2 Visibility of micro AVM components on two types of MRA images
Full size table
Fig. 3
figure 3

DSA and MRA images of a 22-year-old woman with a micro AVM. Anteroposterior (a) and lateral (b) projection images of DSA from the left vertebral artery at late arterial phase reveals a micro AVM nidus (arrow) fed by a branch of the posterior cerebral artery. A micro AVM nidus is not visualized on TOF MRA (c) probably due to a hyperintense hematoma, while a tiny nidus (arrow) is visualized on silent MRA (d). According to both the observers, the nidus was judged as Grade 1 for TOF MRA (c) and Grade 5 for Silent MRA (d)

Full size image
Table 3 Accuracy of each MRA with respect to the three parameters (size, eloquence, deep drainer) and the Spetzler–Martin Grade
Full size table

As six micro CAVMs were not visualized on TOF MRA images, the exact size of the micro CAVMs was not able to measure. Apart from them, in 12 CAVMs, TOF-MRA did not demonstrate their drainers, which led to false classification in Spetzler–Martin grade. By contrast, silent MRA could not provide the correct information about eloquence in two small CAVMs on the source images. In addition, silent MRA could not depict their drainers in six cases with micro CAVM.

Discussion

This study revealed that the sensitivity for the detection of CAVMs and the visibility for nidus and draining veins were higher for silent MRA than TOF MRA. In addition, micro CAVMs, which were defined as a nidus size of less than 10 mm, were detected more effectively for silent MRA than TOF MRA.

TOF-MRA is sometimes insufficient to evaluate certain lesions because it has some characteristic artifacts [13, 14, 26]. Since silent MRA does not depend on the inflow effect into the imaging slab like TOF-MRA, it should be suitable to visualize complex flow in CAVMs. In addition, a high signal due to a hemorrhage is canceled out on silent MRA images owing to subtraction of the images acquired by labelling from those of the control. These effects must have affected our results. In ASL scans, without AV shunt lesions, such false positive lesions cannot be found because T1 dephasing of labeled water is shorter than the capillary transit time.

In the last decade, ASL-based MRI has provided abundant and practical information regarding cerebral AV shunt diseases [18,19,20,21, 23,24,25,26]. Recent development of using pseudocontinous spin labeling, fast imaging acquisitions, and three-dimensional radial sampling trajectories have enabled ASL-based MRA. Schubert et al. demonstrated the excellent sensitivity and specificity of 4D ASL MRA in detecting AV shunt lesions [24]. However, TE of 4D MRA is not as short as that of silent MRA. The blood flow signal indicting complex or turbulent flow in CAVMs could disappear in a short time. Therefore, 4D MRA may not provide clear visualization of nidus or draining veins, compared to silent MRA in which almost 0 TE can minimize phase dispersion in the labeled blood flow signal and reduce magnetic susceptibility artifacts.

Silent MRA may have several merits in neurosurgical clinical practice. Silent MRA has been reported to be effective to delineate aneurysms with stenting [27]. Based on our results, there may be several advantages in patients with CAVM. First, it may be a good screening method for cerebral AV shunt diseases. TOF-MRA cannot depict micro AVMs properly in many cases [30]. Actually, the sensitivity of TOF-MRA for detecting micro AVMs was only 40% in our study, while that of silent MRA was 100%. In specific diseases like hereditary hemorrhagic telangiectasia (HHT), micro or small AVMs are frequent [31]. Silent MRA may be useful for the screening tool. Second, silent MRA may be helpful for following up unstable AVMs. Certain AVMs dynamically change their structures on the venous side, such as varix formation and stenosis of the venous portion, which are risk factors for rupture and require an aggressive surgical policy [4,5,6,7,8,9]. Third, based on the relatively satisfactory accuracy of Spetzler–Martin grade assessment by silent MRA (around 80% in our result), it may help evaluate surgical risk of CAVM instead of other invasive modalities.

In the present study, there were several limitations. First, it may have included unforeseen bias because of its retrospective design. Second, no other ASL-based MRA techniques were included. Recently, such modalities have been reported to be effective in examining CAVMs [18,19,20,21, 23,24,25,26] Further comparative studies with other ASL-based MRA may be needed. Third, there was a difference in the spatial resolution between the two techniques. The scan time of the techniques was clinically acceptable. Fourth, our study population was relatively small. Further studies with a larger number of patients are needed to clarify the role of the technique in the clinical setting.

In conclusion, CAVM components, especially nidus and drainers, and micro AVM were visualized more clearly by silent MRA than by TOF-MRA. Silent MRA may be an effective screening and follow-up modality for CAVMs. Further studies are required to determine the clinical role of this technique.