Vascular and Interventional Radiology
Original Research
Time-Resolved MR Angiography in the Evaluation of Central Thoracic Venous Occlusive Disease
Audio Available | 
Share
OBJECTIVE. The objective of our study was to assess the feasibility and diagnostic performance of time-resolved MR angiography (MRA) in the evaluation of central thoracic venous occlusive disease and to compare time-resolved MRA with conventional MRA and catheter angiography.
MATERIALS AND METHODS. Twenty patients (eight women and 12 men; age range, 19–74 years) with suspected central thoracic venous occlusive disease underwent time-resolved MRA using time-resolved angiography with interleaved stochastic trajectories (TWIST) and parallel acquisition, followed by conventional MRA. Catheter angiography was performed within 1–14 days after MRA and was available for a total of 60 segments for correlation. Time-resolved and conventional MRA images were evaluated in separate reading sessions by two independent radiologists for image quality and level of confidence and degree of venoocclusive disease. The interobserver and intermodality agreement, sensitivity, and specificity were calculated using catheter angiography as the standard of reference.
RESULTS. Time-resolved MRA resulted in diagnostic-quality images that did not differ significantly in quality compared with conventional MRA. Thirty-one segmental venous stenoses were identified. The kappa coefficient revealed moderate intermodality agreement (κ = 0.54; 95% CI, 0.32–0.76) between time-resolved MRA and conventional MRA. When compared with catheter angiography, the sensitivity and specificity for the diagnosis of significant stenosis (≥ 70%) were 87.5% and 68% for time-resolved MRA and 90% and 90% for conventional MRA, respectively.
CONCLUSION. Time-resolved MRA, as described in this study, has the potential to be used as an initial and screening diagnostic tool obviating conventional MRA and its associated higher contrast dose in normal and near-normal examinations. However, because of its relatively lower specificity, adjunct use of conventional MRA is still required for accurate grading of venous occlusive disease.
Keywords: central thoracic veins, hemodynamics, low contrast dose, nephrogenic systemic fibrosis, time-resolved MR angiography, TWIST
| Introduction |   |
|---|
Occlusive disease of the central thoracic veins is commonly seen in patients with malignancy or coagulopathy or in those with long-term use of indwelling central venous catheters for hyperalimentation, chemotherapy, or hemodialysis [1–3]. Imaging plays an important role in detecting the site and extent of occlusion and determining possible causes of occlusion before therapy is initiated. Although catheter angiography is considered the standard of reference for the evaluation of the central veins [4], contrast-enhanced MR angiography (CE-MRA) has been used to good effect in patients with thoracic venous occlusive disease [5–8].
Real-time visualization of the first pass of a contrast bolus was a unique feature of catheter angiography in the past in comparison with static imaging techniques such as CT angiography or MRA. However, because of advances in fast imaging tools, including the introduction of parallel acquisition [9–11] and sparse k-space sampling methods such as time-resolved angiography with interleaved stochastic trajectories (TWIST) [12, 13], time-resolved MRA techniques can now generate high-frame-rate 3D MRA images that allow visualization of the temporal dynamics of blood flow. Another potential advantage of time-resolved MRA is the requirement for substantially less gadolinium than conventional CE-MRA. Considering recent reports regarding gadolinium-based contrast agents as a causative factor in the development of nephrogenic systemic fibrosis (NSF) [14–16], it seems logical to minimize the amount of gadolinium administered to patients, particularly to high-risk populations such as patients with central venous disease, many of whom have underlying renal disease.
Therefore, the purpose of our study was to evaluate the feasibility and diagnostic performance of a 3D time-resolved MRA protocol to evaluate the central thoracic veins and their entire course of inflow and outflow in patients with suspected central venous occlusive disease. The results were compared with conventional MRA and catheter angiography when available.
| Materials and Methods | |
|---|
Twenty consecutive patients (eight women and 12 men; age range, 19–74 years) with suspected occlusive disease of the central thoracic veins were enrolled in this retrospective study. The clinical indications included evaluation of upper extremity or facial swelling (n = 15), known thoracic neoplasia (n = 3), and possible vasculitis in patients with connective tissue disease (n = 2). Ten patients had chronic renal insufficiency with in-place upper extremity dialysis access or a history of central vein cannulation. All patients had correlative imaging including concurrent conventional CE-MRA (n = 20) and follow-up catheter angiography within 1–14 days after MR examination (n = 10). Our HIPAA-compliant study was performed in accordance with institutional review board guidelines under an approved protocol and waiver of informed consent.
All MR examinations were performed on a 32-channel 1.5-T whole-body MR scanner (Magnetom Avanto, Siemens Medical Solutions) equipped with a high-performance gradient system. Patients were placed on the scanner table in the supine position and were advanced head first with their arms at their sides.
For signal reception, a combination of 24 coil elements was used to cover a maximum field of view (FOV) of 500 mm encompassing the entire thorax and the upper extremities. Two six-element body array coils were placed anteriorly and were combined with 12 coil elements from the posterior integrated multichannel spine array coil. If a specific arm had been implicated clinically as the most suspect for venous obstruction, a 20-gauge IV line was sited in the contralateral arm for subsequent contrast injection. If no lateralizing symptoms were present, the IV line was sited by default in the right arm.
Time-resolved MRA was implemented in the coronal plane using a 3D fast gradient-recalled echo (GRE) sequence with the following parameters: TR/TE, 2/0; flip angle, 20°; sampling bandwidth, 1,000 Hz/pixel; FOV, 500 × 458 mm; matrix, 384 × 342; 10–14 partitions; and 1.3 × 1.3 (in-plane) × 7 mm3 voxels. Generalized auto calibrating partially parallel acquisition (GRAPPA) with an acceleration factor of 2 [9] was used, as was TWIST, a recent view-sharing technique that undersamples the periphery of k-space depending on the radial distance of the center of k-space [12, 13]. By combining GRAPPA and TWIST, the resulting 3D data sets were acquired with a temporal resolution of 1.5 seconds per frame for a total of 20 sequential measurements. Automatic in-line subtraction of the 3D data sets was performed online, as were on-axis full-thickness maximum-intensity-projection (MIP) reconstructions by the MR scanner while the time-resolved MRA images were obtained.
A fixed dose of 6 mL of contrast material (gadodiamide [Omniscan, GE Healthcare]) was injected at 2 mL/s, followed by a 30-mL saline flush administered at the same rate using an electronic power injector. Subjects were asked to hold their breath from the start of data acquisition as long as possible and then to breathe gently if necessary.
After IV injection of 0.2 mmol/kg of contrast material (∼ 20 mL), conventional CE-MRA was performed in the coronal plane during suspended respiration using a fast GRE sequence with the following parameters: 2.8/1.1; flip angle, 25°; bandwidth, 610 Hz/pixel; FOV, 500 × 375 mm; matrix, 512 × 330; slice thickness, 1.2 mm; slab thickness, 124 mm; and GRAPPA with an acceleration factor of 3. These settings generated voxel sizes of 1.1 × 1 × 1.2 mm3 during a 20-second acquisition in suspended respiration. Three phases, including one arterial and two venous phases, were acquired.
Ten patients underwent catheter angiography within 1–14 days after MRA examination. After injection of iodinated contrast agent (iohexol [Omnipaque 300, GE Healthcare]), catheter angiography images of the entire venous outflow of the symptomatic arm and central thoracic veins were acquired at 4 frames per second during the contrast agent injection.
Image analysis was performed independently by two radiologists, each of whom had at least 10 years of experience at the time of the study. All angiography images including the temporal online MIP images from time-resolved MRA, partition and MIP images from conventional CE-MRA, and correlative catheter angiography images were available for review on a PACS workstation.
First, time-resolved MRA images were evaluated by the two observers independently. They were informed of each patient's clinical history, but they were blinded to demographic information and other correlative imaging findings. When assessing time-resolved MRA, the observers were asked to note additional dynamic information including the status of inflow and outflow upper extremity veins and the presence of anomalies in the filling order of the arterial and venous system after contrast injection.
Evaluation of correlative studies including concurrent conventional CE-MRA and catheter angiography was performed at least 4 weeks from the time-resolved MRA reading sessions to avoid possible recall bias. These studies were ran domized by the study coordinator and were provided to the same two observers. Both observers were blinded to the patients' names and the findings on time-resolved MRA examinations.
The central venous system was divided into seven segments including the superior vena cava, jugular vein, brachiocephalic vein, and subclavian vein for evaluation of a total of 140 central venous segments. For each venous segment, the severity of venoocclusive disease was evaluated using the following grading scale: 1, normal; 2, mild to moderate stenosis with < 70% luminal narrowing; 3, significant stenosis with luminal narrowing 70–99%; and 4, occlusion. Luminal narrowing was characterized as resulting from either intraluminal thrombosis or extrinsic extraluminal pressure. When two or more stenoses were detected in the same vessel segment, the most severe was used for grading and analysis. Consensus reading of the catheter angiography images was used as the only standard of reference for evaluation of sensitivity and specificity.
Image quality and level of confidence for each venous segment were assessed using the following 4-grade scoring scale: 1, poor image quality and blurring of the venous segment, nonconfident; 2, fair image quality and venous enhancement, somewhat confident; 3, good image quality and enhancement, adequate for confident diagnosis; or 4, excellent image quality and enhancement, highly confident diagnosis. The presence of a pseudoaneurysm or of other incidental vascular findings was recorded for each individual.
A Wilcoxon's signed rank test was used to test for statistical differences between the ratings of image quality and level of confidence of the two observers. The interobserver and intermodality agreement was tested by kappa coefficient and 95% CI for determination of categorized segmental disease [17]. The sensitivity and specificity of MRA in the detection of high-grade stenosis (≥ 70%) were evaluated for each observer using the consensus reading of catheter angiography as the standard of reference.
 View larger version (98K) | Fig. 1A —19-year-old man with antiphospholipid syndrome and history of face swelling and redness. Coronal maximum-intensity-projection images from time-resolved MR angiography (MRA). Using 500-mm field of view enables dynamic depiction of entire thoracic and upper extremity veins with 1.5-second temporal resolution after injection of 6 mL of gadolinium-based contrast agent. Because of occlusion of innominate veins and superior vena cava (SVC), contrast material flows to lateral thoracic and intercostal veins (arrows, frame 3), enhancing azygos vein (arrow, frame 4). Subsequently, retrograde filling of inferior vena cava (arrow, frame 8) and right atrium is seen, followed by filling of pulmonary (arrowhead, frame 10) and systemic (arrowhead, frame 14) circulation. |
 View larger version (170K) | Fig. 1B —19-year-old man with antiphospholipid syndrome and history of face swelling and redness. Conventional contrast-enhanced MRA image shows signal loss along course of previously stented SVC (arrowhead) and multiple collateral veins with retrograde opacification of enlarged azygos vein, consistent with SVC occlusion. |
 View larger version (168K) | Fig. 1C —19-year-old man with antiphospholipid syndrome and history of face swelling and redness. Catheter angiogram confirms occlusion of left innominate vein and stented SVC (arrowhead). |
| Results | |
|---|
All studies were performed successfully and without complications. None of the studies had to be repeated because of technical problems, and all subjects were able to cooperate with breath-holding instructions.
One hundred forty venous segments were available for evaluation. When only time-resolved MRA images were evaluated, the overall average (± SD) score for image quality and level of confidence was 3.19 ± 0.75 and 3.26 ± 0.70 for observers 1 and 2, respectively. The average score for image quality and confidence level for interpretation of conventional CE-MRA was increased to 3.58 ± 0.60 and 3.70 ± 0.54 for observers 1 and 2, respectively (p = 0.08). There was no significant difference in the image quality grading scores between the two observers for time-resolved MRA (p = 0.40) or conventional CE-MRA (p = 0.14).
Time-resolved MRA—In evaluation of time-resolved MRA, central venous stenoses were identified in 31 segments including five segments with < 70% stenosis, 12 segments with 70–99% stenosis, and 14 segmental occlusions. Interobserver agreement for the detection of venous stenosis was good (κ = 0.62; 95% CI, 0.42–0.83). Five patients had no significant stenosis or occlusion.
In one patient with hypercoagulability (antiphospholipid antibody syndrome) (Fig. 1A, 1B, 1C) and a previously inserted stent in the superior vena cava (SVC), the SVC occlusion was apparent on conventional CE-MRA, but the possibility of stent artifact could not be excluded. However, time-resolved MRA revealed the collateral venous circulation and retrograde filling of the right atrium via the inferior vena cava (IVC); these findings indicated the functional occlusion of the SVC. Because this patient presented with face swelling and symptoms of SVC syndrome, catheter angiography was performed for further evaluation, which confirmed occlusion of the stented SVC.
In three patients who had complete SVC occlusion, time-resolved MRA mapped the sequential filling of the collateral circulation, a pattern not apparent on the conventional static MRA images (Figs. 1A, 1B, 1C and 2A, 2B, 2C, 2D). In one patient with a history of synovial sarcoma of the right subclavian fossa and extension to the heart (Fig. 3A, 3B, 3C), MRA showed complete occlusion of the right subclavian vein, right innominate vein, and SVC in addition to a persistent left SVC draining to the coronary sinus.
Conventional CE-MRA—Conventional CE-MRA was concurrently performed after each time-resolved MRA examination, and images were available for all 140 venous segments. Conventional CE-MRA images showed 33 segments with venous disease including 10 segments with < 70% stenosis, 11 segments with 70–99% stenosis, and 12 segmental occlusions. Kappa coefficient revealed excellent overall interobserver agreement (κ = 0.84; 95% CI, 0.67–0.98) for all degrees of venous stenosis, which is significantly higher in comparison with time-resolved MRA. When compared with time-resolved MRA, conventional CE-MRA revealed five additional low-grade stenoses and identified two severe stenoses and two segmental occlusions that had been overestimated by time-resolved MRA. Intermodality agreement was moderate (κ = 0.54; 95% CI, 0.32–0.76) between time-resolved MRA and conventional CE-MRA for the evaluation of venoocclusive disease.
Catheter angiography—Among the 10 patients who underwent both MRA and catheter angiography, 10 vascular segments could not be compared because those particular segments were not shown on catheter angiography, leaving a total of 60 segments for comparative analysis. A consensus reading of catheter angiography for the available segments by both observers indicated disease in 20 of the 60 central venous segments including three segments with < 70% stenosis, nine segments with 70–99% stenosis, and eight segmental occlusions.
When compared with catheter angiography, time-resolved MRA had a sensitivity and specificity for the diagnosis of significant stenosis (≥ 70%) of 87.5% and 68%, respectively. Conventional CE-MRA resulted in a sensitivity and specificity of 90% and 90%, respectively. Table 1 shows the breakdown of the degree of venous stenoses and comparison of MRA techniques with catheter angiography.
| Discussion | |
|---|
Although CE-MRA has been used increasingly for noninvasive evaluation of patients with central thoracic venous occlusive disease [5–8], further improvement in CE-MRA techniques in terms of spatial and temporal resolution and coverage may broaden its acceptance and use as an effective diagnostic and screening tool.
Because of the systemic nature of thromboocclusive disease and associated predisposing factors such as underlying malignancy or end-stage renal disease, central venous occlusive disease may occur repeatedly and sometimes may extend more peripherally into the upper extremity or into more centrally located veins [18, 19]. Therefore, extended coverage has important clinical value, and many clinicians regard evaluation of all central thoracic and upper extremity veins as desirable for further diagnostic and therapeutic decision making [20]. Recognizing this fact, some investigators described a two-step CE-MRA approach to improve the field-of-view coverage for complete coverage of vascular access in patients with dysfunctional hemodialysis access [20, 21].
Multichannel radiofrequency systems with multicoil technology recently became commercially available and were used in this study to improve the performance of the described MRA protocol. Combining multiple high-sensitivity coil elements and multiple independent receiver channels extends the area from which high sensitivity measurements can be obtained. As a result, improvements in the signal-to-noise ratio (SNR), FOV, or a combination of both are possible [22, 23].
 View larger version (113K) | Fig. 2A —46-year-old man with end-stage renal disease and recurrent upper extremity venous thrombosis. Coronal maximum-intensity-projection images from time-resolved MR angiography (MRA). After IV injection of 6 mL of contrast material, time-resolved MRA shows sequential filling of central thoracic veins with temporal resolution of 1.5 seconds. Using 500-mm field of view allows comprehensive evaluation of entire thoracic and upper extremity arterial and venous system in single-station acquisition. Note high-grade focal stenosis of superior vena cava (SVC) (arrows, frames 4 and 5), arteriovenous fistula with aneurysmal venous dilatation in right upper arm (arrowheads, frames 9 and 10), and complete occlusion of right subclavian vein (arrows, frames 13 and 14). |
 View larger version (190K) | Fig. 2B —46-year-old man with end-stage renal disease and recurrent upper extremity venous thrombosis. Conventional contrast-enhanced MRA (B) and catheter angiography (C and D) images confirm complete occlusion of right subclavian vein (arrows, B and C) and significant stenosis of SVC (arrowheads, B and D). |
 View larger version (203K) | Fig. 2C —46-year-old man with end-stage renal disease and recurrent upper extremity venous thrombosis. Conventional contrast-enhanced MRA (B) and catheter angiography (C and D) images confirm complete occlusion of right subclavian vein (arrows, B and C) and significant stenosis of SVC (arrowheads, B and D). |
 View larger version (122K) | Fig. 2D —46-year-old man with end-stage renal disease and recurrent upper extremity venous thrombosis. Conventional contrast-enhanced MRA (B) and catheter angiography (C and D) images confirm complete occlusion of right subclavian vein (arrows, B and C) and significant stenosis of SVC (arrowheads, B and D). |
 View larger version (153K) | Fig. 3A —55-year-old woman with synovial sarcoma involving right clavicular fossa. Coronal maximum-intensity-projection images from time-resolved MR angiography (MRA) show occlusion of right subclavian vein, right innominate vein, and superior vena cava (SVC). Azygos vein fills via collaterals (arrow, frame 3) and left SVC (arrowhead, frames 3 and 4), which then drains to right atrium. Subsequently, pulmonary (arrow, frame 5) and systemic (arrow, frame 7) circulation opacify. |
Our results indicate that the described time-resolved MRA protocol can obtain important clinical information about the anatomic and hemodynamic status of the central veins, providing rapid and comprehensive evaluation of central venous occlusive disease with diagnostic image quality. In addition, the higher SNR gain over a large FOV (500 mm) has enabled us to more efficiently cope with the SNR penalty associated with fast imaging tools such as GRAPPA [9] and TWIST [12, 13], leading to higher protocol flexibility in terms of speed and temporal resolution (1.5 seconds).
The excellent interobserver agreement and acceptable intermodality agreement with conventional CE-MRA of the described time-resolved MRA technique are indicative of its reliability and reproducibility. When compared with catheter angiography, time-resolved MRA showed a sensitivity comparable to conventional CE-MRA for the detection of high-grade stenosis (87.5% vs 90%, respectively). However, interpretation of time-resolved MRA images resulted in a lower specificity (68%) in comparison with the specificity of conventional CE-MRA (90%). The lower specificity of time-resolved MRA is most likely explained by its lower spatial resolution (larger voxel sizes). Additionally, slow contrast flow within severely diseased venous segments may not be adequate for sufficient venous filling and expansion in the short period of time in which time-resolved MRA is performed.
 View larger version (191K) | Fig. 3B —55-year-old woman with synovial sarcoma involving right clavicular fossa. Conventional contrast-enhanced MRA image shows occlusion of right subclavian vein (white arrow), right innominate vein, and SVC in addition to widely patent left SVC (black arrow). Note occlusion of right internal jugular vein (arrowhead). |
 View larger version (208K) | Fig. 3C —55-year-old woman with synovial sarcoma involving right clavicular fossa. Coronal T1 fat-saturated contrast-enhanced MRA image shows large inhomogeneous enhancing lesion at right clavicular fossa (arrows) that is obstructing right thoracic inlet's veins. |
Nevertheless, we believe that time-resolved MRA has several advantages over conventional CE-MRA and has the potential to be used as an effective screening method in patients with central venous occlusive disease.
One of the primary advantages of time-resolved MRA is its high temporal resolution and ability to visualize the dynamics of venous blood flow in a manner similar to that in catheter angiography. In current clinical practice, time-resolved MRA is rapidly emerging as a valuable tool that provides valuable dynamic and functional information in circumstances in which conventional CE-MRA lacks the required temporal resolution. This capability facilitates visualization of the enlarged collateral veins that often accompany chronic venous stenoses and occlusions. This was evident in our study: In three patients with complete occlusion of the SVC, time-resolved MRA mapped the entire collateral circulation and showed the pathway of retrograde filling of the right heart. In one patient, rapid shunting through an arteriovenous fistula was correctly identified with time-resolved MRA.
Second, time-resolved MRA requires a significantly lower contrast dose (6 mL in our study) than conventional MRA (0.2 mmol/kg; ∼ 20 mL) [24]. This difference between the two techniques is of particular importance because gadolinium-based contrast agents have been implicated as a causative factor in the development of NSF in patients with renal insufficiency or failure in a dose-dependent manner [14–16]. The growing population of patients with end-stage renal disease and their increased survival has substantially increased the burden associated with their related complications including diagnoses and management of central venous occlusive disease.
In our study, 10 of the 20 patients (50%) evaluated for central venous occlusive disease had renal insufficiency and thus constituted a population at risk for developing NSF. Both time-resolved MRA and conventional CE-MRA correctly detected that five of these patients (25%) did not have significant central venous disease. If these patients had been studied prospectively, they could potentially have been examined only by time-resolved MRA (only 6 mL of gadolinium contrast agent); the initial time-resolved data sets could have been reviewed by a radiologist for any questionable findings to determine whether an additional conventional CE-MRA examination with its larger contrast dose was necessary.
Third, time-resolved MRA images can be evaluated in less time than CE-MRA images. Although not formally evaluated in our study, the overall interpretation time for time-resolved MRA (20 MIP images) is anticipated to be significantly less than that for high-spatial-resolution MRA images (2–3 sets of 120 images).
We acknowledge several limitations of our study. The study population is small, resulting in relatively low power of the conclusions. Although comparisons were made on a segment-to-segment basis to increase the overall data sample (140 segments), the 95% CIs of our correlative statistics remain wide as a result of the low prevalence of disease. The catheter angiography images used for comparison with MRA images were attained retrospectively, so some patients did not have images of all venous segments available for comparison because the particular images had not been saved in the image archival system. Catheter angiography images were available in only 10 patients, providing a total of 60 segments to be analyzed, and it is important to interpret the measurements of diagnostic performance of this technique in this context. Finally, as implemented, our time-resolved MRA technique is limited by its lack of through-plane resolution. Because through-plane resolution is lower than in-plane resolution, images are best viewed on-axis, where full spatial resolution (in-plane, 1.3 × 1.3 mm2) is maintained. If the MIP images are rotated off-axis, blurring in the slice direction may be evident. Recent advances in k-space trajectories and the reconstruction algorithm may further improve the spatial resolution of time-resolved MRA images to enable us to achieve more isotropic voxel sizes [25, 26].
In conclusion, we successfully implemented a 3D time-resolved MRA protocol over a large FOV to provide a comprehensive assessment of central venous occlusive disease anatomy, collateralization, and hemodynamics. Having a high sensitivity comparable to CE-MRA and needing only a small gadolinium dose, time-resolved MRA has the potential to be used as an initial and screening diagnostic tool, thus obviating conventional MRA in normal or near-normal examinations to significantly reduce the gadolinium dose in this high-risk population. However, because of the relatively lower specificity of time-resolved MRA, adjunct use of conventional CE-MRA seems logical for accurate grading of venous occlusive disease. Further clinical studies are warranted to determine the accuracy of this technique in a broader clinical setting.
Address correspondence to K. Nael ([email protected]).
G. Laub is an employee of Siemens Medical Solutions.
We thank Ali Nael for his contribution to the statistical analyses for this article.
| References | |
|---|
