Feasibility of Single-Shot Whole Thoracic Time-Resolved MR Angiography to Evaluate Patients with Multiple Pulmonary Arteriovenous Malformations

Abstract

Objective

To evaluate the feasibility of single-shot whole thoracic time-resolved MR angiography (TR-MRA) to identify the feeding arteries of pulmonary arteriovenous malformations (PAVMs) and reperfusion of the lesion after embolization in patients with multiple PAVMs.

Materials and Methods

Nine patients (8 females and 1 male; age range, 23–65 years) with a total of 62 PAVMs who underwent percutaneous embolization for multiple PAVMs and were subsequently followed up using TR-MRA and CT obtained within 6 months from each other were retrospectively reviewed. All imaging analyses were performed by two independent readers blinded to clinical information. The visibility of the feeding arteries on maximum intensity projection (MIP) reconstruction and multiplanar reconstruction (MPR) TR-MRA images was evaluated by comparing them to CT as a reference. The accuracy of TR-MRA for diagnosing reperfusion of the PAVM after embolization was assessed in a subgroup with angiographic confirmation. The reliability between the readers in interpreting the TR-MRA results was analyzed using kappa (κ) statistics.

Results

Feeding arteries were visible on the original MIP images of TR-MRA in 82.3% (51/62) and 85.5% (53/62) of readers 1 and 2, respectively. Using the MPR, the rates increased to 93.5% (58/62) and 95.2% (59/62), respectively (κ = 0.760 and 0.792, respectively). Factors for invisibility were the course of feeding arteries in the anteroposterior plane, proximity to large enhancing vessels, adjacency to the chest wall, pulsation of the heart, and small feeding arteries. Thirty-seven PAVMs in five patients had angiographic confirmation of reperfusion status after embolization (32 occlusions and 5 reperfusions). TR-MRA showed 100% (5/5) sensitivity and 100% (32/32, including three cases in which the feeding arteries were not visible on TR-MRA) specificity for both readers.

Conclusion

Single-shot whole thoracic TR-MRA with MPR showed good visibility of the feeding arteries of PAVMs and high accuracy in diagnosing reperfusion after embolization. Single-shot whole thoracic TR-MRA may be a feasible method for the follow-up of patients with multiple PAVMs.

INTRODUCTION

Pulmonary arteriovenous malformations (PAVMs) have a right-to-left shunt that can cause life-threatening complications such as stroke or brain abscesses [1]. Transcatheter embolization is the treatment of choice for PAVM, and a high technical success rate has been reported [2,3,4,5]. However, even after successful embolization, PAVM patency due to recanalization or reperfusion is emerging as an important issue [6,7,8] because it can cause neurologic complications [9]. Therefore, an accurate, noninvasive, and easily accessible follow-up modality is required to monitor treatment failure. CT is used as the standard follow-up modality [9,10]; however, the widely used 70% criteria for venous sac or draining vein reduction have low specificity issues [11,12,13].

In recent years, time-resolved MR angiography (TR-MRA) has shown potential as a novel follow-up modality after PAVM treatment [14,15,16]. This technique demonstrates the flow dynamics of a PAVM without radiation exposure. However, previous TR-MRA studies mainly focused on one PAVM per single contrast injection [17]. Their studies showed limited efficacy in applying this modality to multiple PAVMs due to limited acquisition volume or failure to match CT [14,17]. However, the authors postulated that if single-shot whole thoracic TR-MRA was performed based on a 3T MR system with a high signal-to-noise ratio (SNR), TR-MRA could be used for multiple PAVMs when each lesion was matched 1-to-1 with CT. Therefore, this study aimed to investigate whether single-shot whole thoracic TR-MRA could be used to identify the feeding arteries of the PAVM and reperfusion of the lesion after embolization in patients with multiple PAVMs.

MATERIALS AND METHODS

Study Design and Patient Selection

This retrospective study was approved by the Institutional Review Board of the Kyungpook National University Hospital (IRB No. 2019-12-026). Patients who underwent percutaneous embolization for PAVM between August 2008 and March 2021 were reviewed. To evaluate the efficacy of TR-MRA in the follow-up of multiple PAVMs, eligible patients were identified using the following inclusion criteria: 1) patients with multiple (≥ 2) PAVMs, 2) patients treated with percutaneous embolization therapy using coils or vascular plug, and 3) post-embolization imaging follow-up with CT and TR-MRA performed within 6 months of each other. A total of nine patients (8 females and 1 male; mean age, 47.3 years; age range, 23–65 years) with 62 PAVMs met the criteria (Fig. 1). Electronic medical records were reviewed for data analysis to evaluate clinical history, physical examination, and characteristics of PAVM.

Fig. 1. Flow chart summarizing patient enrollment according to study eligibility criteria.

n = indicates the number of PAVMs, PAVM = pulmonary arteriovenous malformation, TR-MRA = time-resolved MR angiography

CT and Single-Shot Whole Thoracic TR-MRA Techniques

Initial and follow-up CT before and after endovascular embolization was performed mainly using contrast-enhanced CT with multidetector-row scanners. For the examination, 80–100 mL of contrast medium was injected intravenously at a rate of 1.5–2 mL/s. CT images of the area of interest were reconstructed with a 2.5-mm slice thickness in transverse coronal orientation. Follow-up CT was performed at approximately 6 and 12 months and every 2–3 years after PAVM embolization.

All TR-MRAs were performed using a 3T MR system (Signa Architect; GE Healthcare) with a 16-channel cardiac coil. Each acquisition slab was set to maximally cover the entire lung, and images were taken based on a three-dimensional (3D) T1-weighted gradient echo sequence after injection of 0.15 mmol/kg of gadolinium-based contrast (ProHance; Bracco Imaging) at a flow rate of 2 mL/s and saline flushing after breath-hold. The acquisition parameters were as follows: repetition time, 4 ms; echo time, 1 ms; flip angle, 40°; field of view (FOV), 370 mm; matrix, 320 × 192; slice thickness, 5 mm interpolated to 2.5 mm. We acquired 56–71 slices to cover the thickness of the slab, which included the entire lung. Consequently, a voxel size of 1.2 × 1.9 × 2.5 mm³ and temporal resolutions between 1.6 and 1.8 seconds were achieved, and 25 frames of 3D volume were acquired successively. Each frame was reconstructed using a maximum intensity projection (MIP) algorithm.

Image Analysis and Interpretation

All imaging analyses were performed by two independent and experienced cardiovascular radiologists. They were not involved in the clinical diagnosis of PAVMs or follow-up imaging. The outcomes of PAVM embolization, including data and clinical information, were not provided for blinded analysis.

The location, complexity (simple vs. complex), and changes in vessel diameter on preprocedural and postprocedural CT (including feeding artery, venous sac, and draining vein) of PAVMs were reviewed. A simple PAVM is characterized by one or more afferent feeding arteries that originate from a single segmental pulmonary artery, while a complex PAVM is defined as multiple afferent feeding arteries that originate from several segmental arteries [18]. The embolic materials used for embolization and the embolization level (feeding artery embolization vs. venous sac embolization) were also reviewed.

The visibility of the feeding arteries proximal to embolic materials on single-shot whole thoracic TR-MRA was evaluated for all PAVMs by correlating TR-MRA with CT images as a reference. All feeding arteries were identified by comparing preprocedural and postprocedural CT images. The feeding arteries were then examined using TR-MRA images. The search was first performed using coronal whole thoracic MIP TR-MRA imaging, which was subtracted and automatically reconstructed, and subsequently using multiplanar reconstruction (MPR) TR-MRA imaging on a 3D workstation (AW server 3.2; GE Healthcare) (Fig. 2). If the feeding artery proximal to the embolic material could be found in TR-MRA, the reader recorded a positive match. If the feeding arteries were not visible on TR-MRA, the reasons for invisibility were analyzed and reported in consensus between the two readers.

Fig. 2. A 40-year-old female patient underwent coil and vascular plug embolization for three out of four PAVMs located in the left lingular division of the upper lobe and lower lobe.

A. The MIP CT image shows the anteroposterior location of the PAVM. Feeding arteries are noted (arrows). B. Multiplanar reconstructed CT image of 6 months of follow-up. Feeding arteries are noted (arrows). C. The original MIP-reconstructed TR-MRA demonstrates that the anteroposteriorly located PAVM is obscured by the left inter-lobar pulmonary artery (dashed circle). D. After separating the overlapped structure by adjusting the angle through the multiplanar reconstruction of TR-MRA, the matched feeding artery (arrows) is clearly depicted on CT (B). One untreated small PAVM is clearly identified in the whole thoracic TR-MRA (arrow in C). MIP = maximum intensity projection, PAVM = pulmonary arteriovenous malformation, TR-MRA = time-resolved MR angiography

The diagnostic accuracy of TR-MRA in the identification of reperfusion after embolization was evaluated in angiographically confirmed cases. During multiple PAVM treatments with multiple sessions, previously treated lesions were checked by angiography. Segmental or selective pulmonary angiographic images of previously treated lesions were used for angiographically confirmed case analysis. Reperfusion of PAVM in TR-MRA was defined as the simultaneous enhancement or visualization of a venous sac and draining vein in the pulmonary arterial phase [14,15,16].

Inter-reader reliability was assessed using kappa (κ) statistics. Medcalc version 19.4.1 (Medcalc software Ltd.) was used for the analysis.

RESULTS

A total of 62 PAVMs in nine patients (2–20 lesions per patient) were analyzed. The characteristics of the patients, PAVMs, embolization procedures, and follow-up are summarized in Table 1.

Table 1. Characteristics of Patients, PAVMs, Embolization Procedures, and Follow-Up.

Parameters			Value
Patient parameter (n = 9)
	Male, female		1 (11.1), 8 (88.9)
	Mean age (range), year		47.3 (23–65)
	Symptomatic		5 (55.6)
	Number of PAVMs per patient
		2–5	5
		> 5	4
PAVM parameter (n = 62)
	RUL, RML, RLL		9 (14.5), 15 (24.2), 12 (19.3)
	LUL, LLL		10 (16.1), 16 (25.8)
	Simple, complex		60 (96.8), 2 (3.2)
	Mean preprocedure FAD (range), mm		3.01 (1.50–4.90)
	Mean postprocedure FAD (range), mm		2.01 (0.89–4.16)
	Postprocedure FAD categories, mm
		< 1.5	5
		≥ 1.5, < 2	36
		≥ 2, < 3	16
		≥ 3	5
Embolization parameter (n = 62)
	FAE, VSE		46 (74.2),16 (25.8)
	Coils, AVP, AVP + coils		26 (41.9), 35 (56.5), 1 (1.6)
Follow-up
	Mean time from embolization to TR-MRA (range), day		637.5 (4–1567)
	Mean time between CT and TR-MRA (range), day		98 (1–182)

Data are number of patients or PAVMs with percentage in parentheses, unless specified otherwise. AVP = Amplatzer vascular plug, FAD = feeding artery diameter, FAE = feeding artery embolization, LLL = left lower lobe, LUL = left upper lobe, PAVM = pulmonary arteriovenous malformation, RLL = right lower lobe, RML = right middle lobe, RUL = right upper lobe, TR-MRA = time-resolved MR angiography, VSE = venous sac embolization

The visibility of the feeding arteries in TR-MRA assessed by matching TR-MRA with follow-up CT as reference is summarized in Table 2. In comparison, the matching positive (i.e., visibility of feeding arteries on TR-MRA) rates between the original MIP-reconstructed TR-MRA and CT were 51/62 (82.3%) and 53/62 (85.5%) for reader 1 and reader 2, respectively. When each reader analyzed the MPR TR-MRA images on a 3D workstation, the matching positivity increased to 58/62 (93.5%) and 59/62 (95.2%), respectively. The κ between the two readers were 0.760 and 0.792 for the original MIP and MPR images, respectively.

Table 2. Visibility of Feeding Arteries on TR-MRA Compared with CT as the Reference.

	Original MIP TR-MRA (%)		MPR TR-MRA (%)
	Matching + (i.e., Visible)	Matching −	Matching + (i.e., Visible)	Matching −
Reader 1	51 (82.3)	11 (17.7)	58 (93.5)	4 (6.5)
Reader 2	53 (85.5)	9 (14.5)	59 (95.2)	3 (4.8)
Average	52 (83.9)	10 (16.1)	58.5 (94.4)	3.5 (5.6)

Matching + and - as compared with CT indicate visibility and invisibility, respectively, of feeding arteries on TR-MRA. MIP = maximum intensity projection, MPR = multiplanar reconstruction, TR-MRA = time-resolved MR angiography

The factors for the invisibility of feeding arteries in TR-MRA are summarized in Table 3, and examples are shown in Figures 3 and 4.

Table 3. Factors for Invisibility of Feeding Arteries on TR-MRA.

	Original MIP TR-MRA		MPR TR-MRA
	Reader 1 (n = 62)	Reader 2 (n = 62)	Reader 1 (n = 62)	Reader 2 (n = 62)
Feeding artery course on anteroposterior plane	7 (11.3)	6 (9.7)	-	-
Obscured by normal enhancing large vessels (BCV, SVC, right heart, and other PAs)	6 (9.7)	6 (9.7)	-	-
Location adjacent to chest wall	1 (1.6)	1 (1.6)	-	-
Pulsation of heart	1 (1.6)	1 (1.6)	1 (1.6)	1 (1.6)
Small diameter feeding artery	3 (4.8)	2 (3.2)	3 (4.8)	2 (3.2)
Total	11 (17.7)	9 (14.5)	4 (6.6)	3 (4.8)

Data are number of pulmonary arteriovenous malformations with the percentage in parentheses. BCV = brachiocephalic vein, MIP = maximum intensity projection, MPR = multiplanar reconstruction, PA = pulmonary artery, SVC = superior vena cava, TR-MRA = time-resolved MR angiography

Fig. 3. A 49-year-old female patient underwent coil and vascular plug embolization for bilateral 12 pulmonary arteriovenous malformations.

Eleven out of 12 feeding arteries on CT were identified using the original TR-MRA (n = 10) and additionally in multiplanar reconstruction (n = 1).

A. Selective angiography at the time of the procedure shows the feeding artery (arrows) and the segmental artery from which it arises. B. The follow-up CT with maximum intensity projection revealed a markedly decreased diameter of the feeding artery (1.31 mm) (arrow) that is difficult to measure even with CT. C. The multiplanar reconstructed TR-MRA failed to visualize the small feeding artery closest to the embolic material due to limited spatial resolution (dashed circle). TR-MRA = time-resolved MR angiography

Fig. 4. A 44-year-old female patient underwent coil and vascular plug embolization for bilateral 20 PAVMs.

Of the 20 PAVMs, 19 feeding arteries were shown in TR-MRA, matching the findings in CT (in original TR-MRA in 14 and in reconstructed TR-MRA in an additional 5).

A. The chest X-ray showed multiple embolic materials in the bilateral lung fields, including the pericardiac location (dashed circle). B. Maximum intensity projection CT image shows a small feeding artery (arrows). C, D. The matching of the feeding artery (arrow) between CT and TR-MRA failed in both the original (C) and multiplanar reconstruction (D) TR-MRA. The cause of the blurred image was postulated as a heartbeat-induced motion artifact (dashed circles in C and D). PAVM = pulmonary arteriovenous malformation, TR-MRA = time-resolved MR angiography

Regarding the diagnostic accuracy of identifying reperfusion of PAVMs after embolization, 37 PAVMs in five patients were confirmed angiographically during multiple treatment sessions (32 occlusions and 5 reperfusion). Both readers demonstrated 100% (5/5) sensitivity and 100% (32/32, including three cases in which the feeding arteries were not visible on TR-MRA) specificity on TR-MRA to diagnose reperfusion of PAVMs (κ, 1).

DISCUSSION

To evaluate the outcomes of PAVM embolization in TR-MRA, finding each feeding artery compared to preprocedural CT is essential. In the present study, single-shot TR-MRA for multiple PAVMs showed good visibility of the feeding arteries compared to CT, which further improved with the use of the MPR technique. The diagnostic accuracy of TR-MRA for reperfusion after PAVM embolization was high.

TR-MRA has demonstrated high diagnostic accuracy (sensitivity and specificity > 90%) as a follow-up modality after PAVM embolization [14]. This modality has several benefits compared with static CT imaging, such as being radiation-free and providing flow information [14,17]. The hemodynamic characteristics of untreated or reperfused PAVMs in TR-MRA include early venous return through the fistula, which simultaneously visualizes the pulmonary vein in the pulmonary arterial phase. Metallic artifacts from embolic materials (especially metallic coils over vascular plugs) obscure changes in draining vein diameter during CT evaluation. In particular, tightly packed coil nests create a large area of metallic artifacts on CT [4,14]. Minimal blurring artifacts can also be seen on TR-MRA; however, regarding metallic artifacts, TR-MRA showed better resolution for detecting non-shrinkage pulmonary veins or venous sacs, especially in small-sized vessels, due to high SNRs on CT images.

However, there is limited evidence regarding the application of TR-MRA in the follow-up of multiple PAVMs. According to a study by Kawai et al. [17], localized lesions on MRI and CT in a 1-to-1 manner were difficult and required source images of TR-MRA, and multiplanar reconstructed CT images were necessary. However, the incidence of multiple PAVMs is approximately 36% and is particularly high in patients with hereditary hemorrhagic telangiectasia [19,20]. In this study, single-shot whole thoracic TR-MRA was performed using a 3T machine with higher SNRs and large FOV over 1.5T MRI while preserving spatial and temporal resolutions [14]. Furthermore, the MPR showed a high feeding artery matching ratio and good diagnostic accuracy for reperfusion.

There are remaining concerns over feeding arteries that are invisible in TR-MRA. In the present study, the MPR technique was helpful in such cases. However, inter-reader mismatches in pericardiac location and small feeding arteries (< 1.5 mm in diameter) remained even after MPR analysis. These issues have been discussed as limitations of TR-MRA in previous studies [14,17].

Increasing the spatial resolution while maintaining the temporal resolution is the most important factor in reducing non-matching cases caused by a small feeding artery. Hamamoto et al. [21] suggested an ultra-short echo time MRI for PAVM surveillance. Using respiratory gating and ultra-short echo time, they showed high spatial resolution (slice thickness of 1 mm without reconstruction), temporal resolution, and reduced motion artifacts. However, there are some drawbacks, including prolonged acquisition time (7–10 minutes) and lack of hemodynamic information compared to TR-MRA. Ultra-short echo time MR angiography using arterial spin labeling attempted in another study [22] obtained hemodynamic images of treated PAVMs with a high level of temporal resolution (300 ms) and showed strong susceptibility artifacts from the metallic coils. However, it also has the disadvantages of limited spatial resolution compared with CT, small acquisition volume, and long acquisition time (10–12 minutes).

There has been a steady progression of new technologies to improve the spatial and temporal resolution of MR. Compressed sensing is a method to accelerate MRI acquisition by acquiring fewer data through undersampling of k-space [23]. This has the potential to mitigate the time-intensiveness of TR-MRA by shortening the temporal footprint, which is considered a limitation of the view-sharing technique applied in modern TR-MRA [24]. In thoracic imaging, to decrease motion sensitivity and aliasing artifacts, radial sampling of the k-space combined with parallel imaging was introduced in dynamic volumetric MRI [25]. Furthermore, highly constrained projection reconstruction (HYPR) or hybrid HYPR combined with a phase contrast scan or time-of-flight scan has shown improved spatial and temporal resolution [26,27,28]. Furthermore, recent deep learning reconstruction algorithms have shortened the scan and reconstruction time of MR and have improved the temporal resolution in TR-MRA [29,30]. These technical developments are expected to improve the spatial and temporal resolutions of TR-MRA by reducing imaging artifacts.

Currently, non-enhanced CT is the recommended follow-up modality after PAVM embolization [9,10,13]. However, the widely used 70% vein size reduction criteria have low specificity [12,15]. A recent angiographically confirmed case series showed a gray zone in the indeterminate area of CT vein size reduction criteria [12]. The diagnosis of reperfusion is challenging, particularly for multiple PAVMs. The authors recommend that a diagnostic approach using both CT and TR-MRA is necessary, particularly for multiple PAVMs. If follow-up CT using the 70% criteria demonstrates ambiguous results, TR-MRA can be used as an auxiliary tool for a more accurate diagnosis.

This study has some limitations. First, it was a retrospective study. Second, the interval between CT and TR-MRA varied among study patients. The possibility of the new development of PAVMs in the time between CT and TR-MRA may not be completely excluded. More investigation is warranted to establish an optimal follow-up protocol for PAVM outcomes.

In conclusion, single-shot whole thoracic TR-MRA showed good visibility of the feeding arteries of PAVMs compared with CT, which further improved using MPR technique, and high accuracy for diagnosing reperfusion after embolization. Single-shot whole thoracic TR-MRA may be a feasible method for the follow-up of patients with multiple PAVMs.

Availability of Data and Material

The datasets generated or analyzed during the study are available from the corresponding author on reasonable request.