Time-Resolved, High-Resolution Contrast-Enhanced MR Angiography of Dialysis Shunts Using the CENTRA Keyhole Technique With Parallel Imaging

Transcription

1 JOURNAL OF MAGNETIC RESONANCE IMAGING 25: (2007) Original Research Time-Resolved, High-Resolution Contrast-Enhanced MR Angiography of Dialysis Shunts Using the CENTRA Keyhole Technique With Parallel Imaging Katja A. Mende, MD, 1 * Johannes M. Froehlich, PhD, 1 Constantin von Weymarn, PhD, 1 Romhild Hoogeveen, PhD, 3 Thomas Kistler, MD, 2 Christoph L. Zollikofer, MD, 1 and Klaus U. Wentz, MD 1,4 Purpose: To evaluate the use of a dynamic keyhole magnetic resonance angiography (MRA) sequence combined with sensitivity encoding (SENSE) for hemodialysis shunts, because surveillance with conventional contrast-enhanced MRA (CE-MRA) is limited by its low temporal resolution, resulting in arteriovenous overlay. Materials and Methods: A total of 12 patients with Brescia-Cimino shunts were investigated prospectively using the new technique. During the contrast passage (gadoterate, Gd-DOTA) a series of five to nine dynamic central k-space measurements (10% for upper-arm shunt, 25% for lower-arm shunt) followed by a full reference data set were acquired. The outer k-space data of the single reference scan were used to complete the dynamic data sets. Results: All studies were diagnostic (17 stenoses, three aneurysms) without complications. The acquisition times for a single dynamic scan of the upper- and lower-arm shunts were 2.2 and 3.2 seconds, respectively, while the reference scan needed 13 and 22.4 seconds, respectively. The dynamic angiograms allowed the differentiation of arterial and venous filling despite a mean peak delay time of only 4.2 seconds in the shoulder region. Image quality qualified in consensus by two experienced readers was rated good in 19 cases and intermediate in five cases with high mean values for signal-to-noise ratios (SNRs) and contrast-to-noise-ratios (CNRs). 1 MR-Research Group, Institute of Radiology, Cantonal Hospital, Winterthur, Switzerland. 2 Division of Nephrology, Department of Internal Medicine, Cantonal Hospital, Winterthur, Switzerland. 3 Philips Medical Systems, Best, The Netherlands. 4 Department of Radiology and MicroTherapy, University Witten-Herdecke, Witten-Herdecke, Germany. *Address reprint requests to: K.A.M., Department of Radiology, Cantonal Hospital, Brauerstrasse 15, CH-8401 Winterthur, Switzerland. katja@mende.ch Received January 20, 2006; Accepted October 30, DOI /jmri Published online 7 March 2007 in Wiley InterScience ( wiley.com). Conclusion: We have successfully implemented a fast, dynamic, CE-MRA technique with CE timing robust angiography (CENTRA) keyhole and SENSE in clinical routine. High spatial and temporal resolution improve the diagnostics of dialysis shunts and allow the assessment of detailed, dynamic, four-dimensional (4D) information. Key Words: contrast-enhanced MR angiography; keyhole; parallel imaging; hemodialysis shunts; gadoterate J. Magn. Reson. Imaging 2007;25: Wiley-Liss, Inc. THE MOST IMPORTANT REQUIREMENT for long-term hemodialysis is an adequate functioning of the arteriovenous fistulas (AVF) or arteriovenous grafts (AVG). The major cause of shunt failure is stenosis and consecutive thrombotic occlusion (1). Therefore, the early detection and treatment of shunt stenoses is most essential (2,3). Currently, vascular complications, stenoses, and occlusions are predominantly detected and graded by Doppler-ultrasound and digital subtraction angiography (DSA). The latter is usually performed with the intention to treat. The major disadvantage of this relatively invasive approach is the exposure to ionizing radiation of both the patient and the investigator, and the need for iodinated contrast media. In addition to the higher hazard of allergic reactions or of thyroid complications, these patients risk a worsening of their residual renal function. Therefore, contrast-enhanced (CE) three-dimensional (3D) magnetic resonance angiography (MRA) might be a potentially favorable diagnostic alternative (nonionizing radiation, improved tolerance of contrast) for patients with suspected shunt failure. Investigation of hemodialysis shunts with MRA is extremely challenging due to high blood flow conditions, ranging from 600 to 1200 ml/minute, according to our experience and the literature (4). Conventional CE- MRA, as reported previously, suffers from the following: insufficient temporal resolution resulting in arteriovenous overlay; difficulties visualizing the venous runoff vessels up to the superior vena cava; and overestimation of stenoses (2,5). A major drawback of time-of Wiley-Liss, Inc. 832

2 CENTRA for CE-MRA of Dialysis Shunts 833 Figure 1. Schematic overview of the principle of keyhole-technique: a predefined portion (in %) of the central region of k y -k z space is repeatedly acquired (first row), followed by a single acquisition of the peripheral k-space corresponding to the reference scan (last scheme on the right). In the reconstruction (second row), the central, dynamic part of k-space is combined with the static periphery of k-space acquired during the reference scan. The corresponding dynamic MIPs of the contrastenhanced MR-angiograms are acquired every 2.2 seconds, demonstrating the high temporal ( seconds) and spatial resolution (third row) of the method. flight and phase-contrast acquisition methods is the frequent occurrence of flow artifacts in regions with disturbed flow (6,7) and their prolonged examination time. Compression techniques such as timed arterial compression (TAC-MRA) show a high spatial resolution but lack temporal resolution (2,8,9). CE-MRA with CE timing robust angiography (CEN- TRA) keyhole and sensitivity encoding (SENSE) promises to provide additional dynamic information while maintaining high spatial resolution. CENTRA keyhole is a combination of the CENTRA technique with keyhole imaging (10). The CENTRA technique is basically a phase-encoding reordering technique (11). It segments k y -k z space into two regions, with the central disk or cylinder acquired first in random order, followed by the periphery of k-space. With the CENTRA technique, high-resolution arterial, venous-free CE-MRA images can be obtained, even though acquisition times are on the order of one to two minutes. In CENTRA keyhole, the central region is then repeated a couple of times, followed by the periphery of k-space (Fig. 1). The periphery of k-space is typically acquired while the contrast agent is recirculating, thereby receiving signal from both arteries and veins. In the reconstruction, the central, dynamic part of k-space is combined with the static periphery of k- space for each dynamic step. The size of the central region can be configured: the smaller the size, the higher the temporal resolution. In practice, a central portion of k-space of 10% to 20% corresponds to a speed-up factor between 10 and 5, respectively. A further increase of temporal and/or spatial resolution may be achieved combining the CENTRA keyhole technique with parallel imaging (e.g., SENSE). SENSE is a technique designed to reduce imaging time by undersampling k-space and recording images simultaneously from multiple imaging coils. The purpose of this prospective clinical study was to evaluate image quality and quantitative contrast-enhancement characteristics of a fast, dynamic, CE-MRA technique with CENTRA keyhole and SENSE, to prove feasibility in patients with hemodialysis shunts. MATERIALS AND METHODS Patients A total of 12 hemodialysis patients (five with upper armand seven with lower-arm Brescia-Cimino shunts) in the Division of Nephrology (Cantonal Hospital, Winterthur, Switzerland) underwent keyhole MRA between January and August Five of our patients were clinically suspected of suffering from shunt dysfunction; the other seven were included for follow-up purposes to image shunt functioning. Exclusion criteria were contraindications to MRI such as the presence of ferromagnetic implants, pacemakers, or known intolerances to gadolinium contrast agents. The study and protocol were approved by our Institutional Review Board. Informed consent was obtained from all 12 patients. MRI The MRI examinations were performed using a 1.5-T clinical MR scanner (Gyroscan-NT INTERA R, software release 9.1.1; Philips Medical Systems, Best, The Netherlands) equipped with high-performance gradient systems (maximal gradient amplitude 30 mt, slew rate 150 T/m/second, maximum field of view (FOV) 530 mm). For improved signal detection, a commerciallyavailable four-element phased-array body coil allowing parallel imaging was used, while the ordinary body coil was used as the transmitter coil.

3 834 Mende et al. postcontrast 3D MRA of the shunt region (either the upper or lower arm); and 5) a complete 3D MRA of the consecutive location before and after injecting a second contrast media bolus. The patients were placed in a supine head-first position with the shunt arm lying alongside the body. For imaging of both the upper and the lower arm, the relevant part was positioned between the two anterior and two posterior flexible arrays of the coil, while the hand was fixed in a sagittal position. A constant distance between both parts of the coil was maintained by using a wooden device (Fig. 2a and b), which was placed along the extremity. Figure 2. Optimal positioning of the patient for imaging of the upper arm (a) and lower arm (b). Note the wooden device that maintains a constant distance between the coil and the extremity. [Color figure can be viewed in the online issue, which is available at The MR angiographic examination consisted of five steps: 1) positioning of the patient; 2) determination of the vessel anatomy by performing maximum intensity projections (MIP) of transverse 2D time-of-flight MRA sequences with a low matrix for the in-plane resolution; 3) acquisition of a timing sequence with a temporal resolution of 1.0 second to determine the time of arrival and transit-time of the contrast material for both the upper and lower arm location; 4) performing of pre- and CE 3D CENTRA Keyhole MRA First a full 3D data set without contrast was acquired, which could be used for subtraction later on. This was followed by a fast, dynamic, high-resolution CE-MRA with a CENTRA keyhole k-space acquisition technique with the parameters given in Table 1. The two singlestation MRAs encompassed the peripheral vascular tree from the aorta to the peripheral arm arteries. This approach was chosen due to high flow conditions. Total scan duration depended mainly on the total number of dynamic keyhole scans (five to nine times, 2.2 or 3.2 seconds) and duration of the reference scan (13.0 or 22.4 seconds). The number of acquired keyhole scans was adapted to the bolus transit-time seen during the timing sequence. This adaptation was performed with the goal of allowing optimal dynamic depiction concomitantly with a standardized contrast enhancement during the acquisition of the reference scan. This methodology, at least theoretically, should allow acquiring the reference scan with still contrastfilled vessels (12). Measurement of the reference scan was obtained in 22.4 seconds for the upper arm and within 13 seconds for the lower arm, respectively. Contrast Administration For contrast media application an intravenous line was placed into an antecubital vein of the contralateral arm (22-G cannula, Vasofix Certo; Braun, Melsungen, Germany). Injection was performed under standardized Table 1 Keyhole MRA Sequence Parameters for the Upper and Lower Arm Parameter Upper arm Lower arm TR/TE/flip-angle (msec/msec/ ) 3.8/1.35/45 3.9/1.43/45 Matrix Reconstruction Acquisition percentage 100% 100% FOV 256 mm 256 mm 368 mm 368 mm Partition thickness 1.5-mm Fourier-interpolated 1.5-mm Fourier-interpolated Number of partitions SENSE Factor 2 in phase direction Factor 2 in phase direction CENTRA Yes Yes Keyhole percentage 10% 25% Keyhole scan duration 2.2 seconds 3.2 seconds Reference scan duration 22.4 seconds with 3D, Fourier-interpolated slices 13.0 seconds with 3D, Fourier-interpolated slices Acquired voxel x3mm mm 3

4 CENTRA for CE-MRA of Dialysis Shunts 835 conditions by using an MRI-compatible power injector (Spectris; Medrad, Indianola, Pennsylvania, USA). Timing was individually assessed for the upper and lower arm with a bolus injection of 3 ml meglumine gadoterate (Gd-DOTA, Dotarem ; Guerbet, Roissy, France) followed by 30 ml of saline solution (NaCl 0.9%) at a flow rate of 2.5 ml/second. The bolus arrival and passage time were assessed using a real-time 2D projection MR- DSA technique (BolusTrak ; Philips Medical Systems) with a temporal resolution of 1 second. For the acquisition of the keyhole MR angiograms the same flow rates and saline flush volumes were applied using two separate single-dose injections (0.2 ml/kg body weight [bw]) of Gd-DOTA for the upper arm and forearm. This resulted in a total dose of 0.4 ml/kg bw of intravenous (i.v.) Gd-DOTA. Image Analysis MIPs of subtracted and nonsubtracted image data sets were evaluated on a workstation (Easy Vision; Philips Medical Systems). Overall image quality and diagnostic value were assessed in consensus by two experienced readers (K.M., K.W.) on a three-point scale (good, intermediate, poor) using the 3D-MIP reconstructions. Good image quality referred to an image that had no layering artifacts, with good discrimination between arteries and veins, with optimal depiction of shunt anatomy, and optimal differentiation of contrast dynamics on the single acquisitions. Intermediate quality referred to the presence of artifacts or low contrast hampering the differentiation of the vessels from the background. Poor image quality referred to severe artifacts, arteriovenous overlays, reduced visibility of vessel segments, difficulties to identify the fistula, or poor overall quality. Signal-to-noise ratios (SNRs), vessel-soft tissue contrast-to-noise-ratios (CNRs) and dynamic enhancement characteristics were evaluated for all patients using the MIP data. Standardized ROIs were placed in the aorta, in the subclavian artery, in the anastomosis, in the venous outflow tract in vicinity to the fistula, in the axillary vein, and finally in the superior caval vein. The soft tissue signal intensities (SIs) were measured within regions of surrounding muscles if possible without overlaying artifacts. Background signal was measured in air laterally to the upper arm, avoiding areas of phase ghosting. CNR and SNR were calculated for each dynamic scan using the following formulas: SNR SI vessel /SD background and CNR (SI vessel SI tissue )/SD background, where SD is standard deviation. These were plotted along the time axis for the various vascular ROIs. Dynamic enhancement characteristics were assessed graphically (Fig. 3). Statistics For descriptive analysis numerical data are presented as mean SD. Figure 3. Example of CNR dynamics for a single patient with an upper arm shunt. Due to high flow conditions, peak delays are extremely short, but may be differentiated due to the high temporal resolution of the keyhole technique. RESULTS Patients All patients (one female, 11 males; mean age 60.3 years, range 29 to 75 years) were successfully examined. The MR examinations were well supported without any adverse effects. All underwent hemodialysis within a maximum of 72 hours after the MR examination. Qualitative Analysis The MIP displayed with high-spatial resolution the entire topography of the feeding arteries, anastomoses, and draining veins from the aorta, over the graft to the palmar arch (Fig. 4). The dynamic angiograms allowed differentiation of arterial and venous filling despite the typically high flow rates of hemodialysis shunts (Fig. 1; third row) (4). The contrast dynamics could be followed demonstrating the fast progressive filling of the vessels and typical first-pass bolus passage with its short maximum peak enhancement in individual vessel segments. Overall image quality for the upper arm was rated as good in N 11 cases, intermediate in N 1, and of poor quality in N 0 cases. For the lower arm, N 8 cases were rated as good, while N 4 were intermediate. All studies were diagnostic. The following pathologies could be assessed: N 2 stenoses of the arterial branch; N 2 within the arteriovenous fistula; N 13 along the venous outflow; and N 3 aneurysms. DSA was performed for further diagnostic clarification or therapeutic reasons in four patients, allowing comparison of both methods in those cases. A stenosis in the proximal cephalic vein was correctly identified with both methods in one case (Fig. 5). In a second patient, a high-grade tandem stenosis with poststenotic aneurysms in the venous outflow tract of the shunt could be reproduced with DSA (Fig. 6). A third patient had a high grade stenosis in the primary inflow tract (radial artery) on MRA with a consecutive collateral inflow tract formation (ulnar artery). A second 50% stenosis in the same patient was visible in the venous segment adjacent to the fistula. The latter was confirmed with DSA, while the arterial stenosis seemed to be occluded in DSA. The weak arterial backflow due to retrograde i.v.

5 836 Mende et al. injection of the contrast agent in DSA, combined with the high-grade stenosis, probably lead to this falsepositive result, with an abrupt breakoff of the vessel in DSA (Fig. 7). The last patient examined both with MRA and DSA presented a 50% stenosis proximal to the fistula point and a high-grade stenosis in the venous outflow tract in both modalities. In most of the patients with upper arm shunts, the arteries of the forearm, which were assessed with a second contrast bolus, showed reduced contrast due to a steal phenomenon. Quantitative Analysis Contrast dynamics with a clear separation of inflow and outflow could be assessed for all patients despite their extremely short arterial and venous passage times. High mean SNR and CNR values with relatively high SDs were assessed in both the arterial and venous branches (Table 2; Fig. 3). Only small intraindividual differences due to flow phenomena or presumably peripheral dilution were noted, whereas larger interindividual differences explain larger SDs. Peak delays between the arterial and venous phases were extremely short (mean 4.2 seconds) when measuring at the axillary level and approached the limits of temporal resolution ( seconds for one single keyhole dynamic) (Fig. 3; Table 2). Relatively high SDs of the background SIs attending a maximum of were especially noted at the level of the upper arm and related to pulsation artifacts from the heart. DISCUSSION Figure 4. Example of a 3D CENTRA keyhole acquisition of a lower arm shunt combined with the subsequent acquisition of the upper arm. Note the extended coverage and optimal contrast in both anatomic regions despite two separate acquisitions. The major result of this study is the applicability of a segmented k-space acquisition technique for MR angiographic imaging of shunt anatomy and pathology in dialysis patients. The combination of advanced k-space sampling techniques (keyhole, CENTRA) not only allows extension of the acquisition time as usual by selectively sampling the contrast-determining k-space lines during the first-pass of the contrast agent, but also to boost temporal resolution down to 2.2 or 3.2 seconds, respectively, for a single dynamic. The peripheral parts of k-space, necessary for spatial resolution, are acquired with the keyhole reference scan. This oc- Figure 5. A relevant stenosis (arrows) in the proximal cephalic vein was identified with 3D CENTRA keyhole MRA (left) and could be confirmed with DSA (right).

6 CENTRA for CE-MRA of Dialysis Shunts 837 Figure 6. Keyhole MRA (left) of a patient with lower arm shunt shows a tandem stenosis (arrows) and poststenotic aneurysms (arrowheads) in the venous outflow tract, which presented an identical appearance within DSA (right). curs rather late, after the acquisition of all dynamics, but still with arterial and venous contrast filling. According to our knowledge, this is the first clinical study in which the feasibility of dynamic 3D CENTRA keyhole Figure 7. Patient presenting high-grade stenosis in the radial artery (arrow) on MRA (left) with a collateral inflow via the ulnar artery (black arrowhead). A second, approximately 50% stenosis is visible in the venous segment adjacent to the fistula (white arrowhead). The latter was confirmed with DSA (right white arrowhead), while the arterial one seems to be occluded in DSA (arrow). CE-MRA has been successfully tested in this particular patient group with extremely high blood-flow conditions. Its decisive advantage is the boosting of temporal resolution while keeping a high spatial resolution over a large FOV. The upper extremities are the most frequently used approach for hemodialysis access in patients with endstage renal failure. Different imaging approaches have been evaluated for diagnostic of upper extremity vasculature (13 16) and dialysis shunts (17 19). Planken et al (17) reported the applicability of 3D MRA with a temporal resolution of 10 seconds, but recorded a rather insufficient spatial resolution (3.1 mm 3 interpolated to 1.5 mm 3 ; FOV , matrix size 304 mm 58 mm, partition thickness 2 mm), explaining their relatively low specificity for stenosis detection of 10% despite their high sensitivity of 100%. Results of a recently published study (19) suggest that conventional CE-MRA techniques, using a rather high spatial resolution (sections 120, overcontiguous 0.55 mm, voxel size 1 1 1,1mm 3, interpolated mm 3,t 32 seconds for the shunt region; inflow tract: sections 95, overcontiguous 0.9 mm, voxel size , interpolated mm 3,t 32 seconds) allow increasing specificity up to 99% while preserving high sensitivity with 97%, positive predictive value (PPV) 96% and negative predictive value (NPV) 99%, despite a reduced temporal resolution of 32 seconds. However, according to our experience, these previously assessed temporal resolutions of 10 or 32 seconds do not allow sufficient separation of the arterial inflow and venous outflow and analysis of complex shunt anatomies with a higher portion of collaterals. As reduced inflow seems to correlate with a higher degree of stenosis (4), the profound analysis of contrast dynamics could become of practical relevance (20). Higher temporal resolution may principally be obtained by various techniques, such as parallel imaging (SENSE), k- space interpolation, half-fourier reduction of matrices or FOVs, partial echo, echo-planar imaging (EPI), et al. Our chosen segmented k-space acquisition technique (keyhole CENTRA), in combination with k-space reor-

7 838 Mende et al. Table 2 Bolus Travel Time From Aorta to Region of Interest, SNRs, and CNRs for All Patients (N 12)* Peak aorta Peak axillary artery Peak shunt Peak axillary vein Peak delay aorta to ROI (seconds) Defined zero point SNR CNR *Note the short interval between arterial and venous bolus with 4.2-second difference (measured in the axillary artery and vein), demonstrating the high temporal resolution of the method. dering (SENSE), not only allows boosting temporal resolution with acquisition times for a single volume of the upper arm or lower arm of 2.2 or 3.2 seconds, respectively, but also permits a high spatial resolution (upper arm: mm; lower arm: mm) of the acquired MR angiograms. It should be noted that the FOV is kept stable and allows visualization of the whole vascular territory between the aortic and palmar arch. The fast dynamic acquisition allowed arteriovenous differentiation in all cases. In case of some overlapping of the two phases, this was probably caused by the relatively long contrast bolus chosen to improve the reference scan, as discussed below. Timeresolved MRA in our case allowed quantifying the peak transit times for all patients and segments. For the whole group of patients transit delay times could be assessed between a mean of 5.86 seconds for the aortic arch and seconds for the axillary vein postinjection, in accordance with the high-flow conditions in these dialysis shunts (4). Potentially, such a fast and less invasive method offers, at least theoretically, a number of benefits with respect to other methods: possible reduction of imaging time, provision of dynamic information, perfusion information, possible reduction of contrast volume (better renal tolerance), reformatting of 3D data in various directions, possibility to assess flow rates or improved differentiation of single phases of first-pass bolus passage (arterial, venous). Various Technical Aspects Bolus Timing Even though a timing sequence was still performed according to our standard protocol, the acquisition of several dynamics, which allowed us to choose the optimal one in postprocessing, definitely reduces the importance of careful bolus timing. Compared to CENTRA alone, keyhole seems to forgive timing errors (21). Keyhole Parameters Besides the conventional imaging acceleration methods (SENSE, CENTRA), the main gain of time is realized by choosing the percentage of keyhole. It is not known whether the chosen percentages are optimal regarding all parameters. The well known SNR loss described for parallel imaging techniques (approximately 30%) does not seem to increase with keyhole, even though this has not been studied systematically in this study. The temporal resolution is increased by a factor of 4 to 10 by CENTRA keyhole and doubled by using SENSE, resulting in a total amplification of temporal resolution by a factor of 8 to 20. This might be slightly at the expense of overall image quality, but offering time-resolved depiction of first-pass contrast passage and differentiating flow phases in hemodialysis patients often suffering from complex vessel pathologies is reasonable. It is important to outline that the speeding-up of the protocol compared to nonkeyhole techniques is realized by preserving spatial resolution and the imaged FOV. Reference Scan The final and prolonged acquisition of the reference scan at the end after the passage of the contrast bolus allows using the residual amounts of contrast during the washout or recirculation phase of contrast possibly improving vessel conspicuity. Faint contrast enhancement during the acquisition of the reference scan leads to an improved visibility of the whole upper extremity vessels on all MIP images as was assessed in preliminary tests. On the other hand, the contrast-filled vessels during measurement of the reference scan at the end of the imaging protocol may contaminate initiallyacquired MIPs, which otherwise would not contain any contrast in the vessels. This might be judged as disturbing, but it seems that the effect may be neglected as the dynamic scans, due to their high CNR during first-pass transit of the contrast bolus, allow differentiating the single vascular phases (arteries, shunt, runoff vessels with veins). Acquisition of the reference scan at the beginning without contrast seems to decrease vessel conspicuity and image quality, as preliminary tests had demonstrated. It remains to be clarified if a reduction of total contrast dose, shortening the bolus itself, and thus also reducing the amount of contrast during the acquisition of the reference scan, still results in similar image quality. Two-Station Technique With our specific two-station protocol, depending on the localization of the shunt, we were able to cover the whole extent from the heart to the fingertips in less than about 10 minutes total examination time. This twostation technique with separate contrast injections for imaging of the upper and lower arm was preferred, as the moving table approach is unfavorable for the upper extremity. In addition, considering the fast blood flow conditions in hemodialysis shunts, such a prolonged

8 CENTRA for CE-MRA of Dialysis Shunts 839 acquisition would necessarily lead to arteriovenous overlay with missing dynamics. Artifacts Some problems were related to the presence of pulsating artifacts from the heart, ghost artifacts, and vessel blurring. Ringing artifacts, which have been described for keyhole (22), were not visible and were probably reduced by CENTRA. Additionally, the coronal FOV only covered about one-half of the thorax with the arm of interest rather placed laterally. This increases the likelihood of foldover artifacts. (Phase encoding was in the right-to-left direction for the upper arm and in the anterior-to-posterior direction for the lower arm) (23). High blood flow conditions through the shunt are probably another source for image deterioration (flow artifacts). Nevertheless, stenoses could be assessed well even in the more difficult runoff vessels despite the increased risk of flow artifacts or other potential flowdependent disturbing factors (displacement artifact). Arguably, the rather late and separate acquisition of the reference scan compared to the dynamic scans could lead to more movement artifacts and thus to a higher sensitivity to motion but this was not noted in this study. This underlines the importance of a good positioning and fixation of the whole arm in such studies. SNR and CNR Compared to our usual CE-MRA protocol, we assessed comparable SI values, but an increased background noise level (9). This which might be related to the keyhole acquisition technique. The influencing factors must still be clarified. It also remains unclear whether keyhole CENTRA translates to reduced vessel conspicuity. SI maps orthogonal to the larger vessels showed a relatively steep slope on both edges with a relative narrow central peak. We were able to visualize the hemodialysis shunts in all cases with high quality, as reported. Even complex venous outflow tracts could be analyzed. Stenoses, occlusions, and aneurysms were correctly identified in four patients who also had consecutive DSA. Contrast dynamics could be visualized in all cases with clear bolus-transit depiction. In most of the cases with upper arm shunts, we saw a reduced depiction of the vessels of the forearm, most likely due to a steal-effect caused by the shunt. As shown, CE-MRA offers the advantage of depicting the whole arterial vasculature of the shunt, while this is not the case in DSA, in which an intravenous retrograde approach is usually preferred. Even the compression techniques, which are frequently used in DSA to improve the visualization of the arterial segment, allow only the evaluation of proximal arterial segments near the shunt. Figure 7 demonstrates a complete occlusion of the radial artery in DSA, while a rest caliber is still visible in CE-MRA, despite the advanced imaging strategy. An additional advantage in addition to this better coverage is the reduced invasiveness compared to DSA. Limitations This feasibility study was performed to implement a segmented k-space acquisition technique for diagnostic purposes in a patient group that could potentially profit from a fast, less invasive, and time-resolved technique. Thus we did not systematically compare intra- or interindividually various diagnostic methods such as DSA, Doppler ultrasound, computed tomography (CT) angiography, or other MRA methods with the suggested new CE-MRA method. Patients were only redirected to other modalities in case of suspected or evident pathology in CE-MRA. Accordingly, prognoses regarding the various endpoints of diagnostic accuracy are lacking. Our results suggest that it would be worth testing the methods under discriminatory conditions. The rather low number of patients is a further limitation when comparing our study results with other studies (17,19,20). The examination protocol and, accordingly, the choice of keyhole parameters must also be further assessed and probably optimized. Other k-space segmentation or undersampling techniques, such as time-resolved imaging of contrast kinetics (TRICKS) from General Electric (24), time resolved echo-sharing angiographic technique (TREAT) from Siemens, or newer algorithms such as the socalled Broad-use Linear Acquisition Speed-up Technique (k-t BLAST) (25), have completely different approaches compared to keyhole. Comparisons of these various approaches and their respective strengths and weaknesses are still lacking. Finally, an improved coil design better adapted to the anatomy of the upper extremities is warranted as well and could further amplify the image quality. In conclusion, we have successfully implemented a 4D fast, dynamic, CE-MRA technique with SENSE and CENTRA keyhole for imaging of hemodialysis shunts. The method combines both high spatial and temporal resolution for the diagnostics of dialysis shunts and allows assessing detailed flow-resolved information. Flow obstructions and detailed vascular anatomy and flow dynamics of the proximal and distal fast-flowing upper extremity shunt vessels could be depicted. The technique may help to plan adequate interventions in a high risk population. Further, it promises to improve the diagnostics in other high-flow vascular regions. In the future, its diagnostic accuracy should be evaluated in prospective comparative studies using DSA as the gold standard. ACKNOWLEDGMENTS We thank Mrs. Lisa Schätzle and Mr. Thorsten Weihrauch for performing the MR studies of the patients. REFERENCES 1. Rotmans JL, Pasterkamp G, Verhagen HJ, Pattynama PM, Blankestijn PJ, Stroes ES. Hemodialysis access graft failure: time to revisit an unmet clinical need? J Nephrol 2005;18: NKF-DOQI clinical practice guidelines for vascular access. National Kidney Foundation-Dialysis Outcomes Quality initiative. Am J Kidney Dis 1997;30(Suppl 3):S150 S191.

9 840 Mende et al. 3. Manninen HI, Kaukanen ET, Ikaheimo R, et al. Brachial arterial access: endovascular treatment of failing Brescia-Cimino hemodialysis fistulas initial success and long-term results. Radiology 2001;218: Pietura R, Janczarek M, Zaluska W et al. Colour Doppler ultrasound assessment of well-functioning mature arteriovenous fistulas for haemodialysis access. Eur J Radiol 2005;55: Han K-M, Duijm LEM, Thelissen GRP, et al. Failing hemodialysis access grafts: evaluation of complete vascular tree with 3D contrast-enhanced MR angiography with high spatial resolution: initial results in 10 patients. Radiology 2003;227: Bos C, Smits JH, Zijlstra JJ, et al. MRA of hemodialysis access grafts and fistulae using selective contrast injection and flow interruption. Magn Reson Med 2001;45: Bilecen D, Aschwanden M, Heidecker HG, Bongartz G. Optimized assessment of hand vascularization on contrast-enhanced MR angiography with a subsystolic continuous compression technique. AJR Am J Roentgenol 2004;182: Wentz KU, Froehlich JM, von Weymarn C, Patak MA, Jenelten R, Zollikofer CL. High-resolution magnetic resonance angiography of hands with timed arterial compression (tac-mra). Lancet 2003; 361: Smits JH, Bos C, Elgersma OE, et al. Hemodialysis access imaging: comparison of flow-interrupted contrast-enhanced MR angiography and digital subtraction angiography. Radiology 2002;225: Boos M, Scheffler K, Haselhorst R, Reese E, Frohlich J, Bongartz GM. Arterial first pass gadolinium-cm dynamics as a function of several intravenous saline flush and Gd volumes. J Magn Reson Imaging 2001;13: van Vaals JJ, Brummer ME, Dixon WT, et al. Keyhole method for accelerating imaging of contrast agent uptake. J Magn Reson Imaging 1993;3: Willinek W, Gieseke J, Conrad R, et al. Randomly segmented central k-space ordering in high-spatial-resolution contrast-enhanced MR angiography of the supraaortic arteries: initial experience. Radiology 2002;225: Connell DA, Koulouris G, Thorn DA, Potter HG. Contrast-enhanced MR angiography of the hand. Radiographics 2002;22: Winterer JT, Ghanem N, Roth M, et al. Diagnosis of the hypothenar hammer syndrome by high-resolution contrast-enhanced MR angiography. Eur Radiol 2002;12: Krause U, Pabst T, Kenn W, Hahn D. High resolution contrast enhanced MR-angiography of the hand arteries: preliminary experiences. Vasa 2002;31: Goldfarb JW, Hochman MG, Kim DS, Edelman RR. Contrast-enhanced MR angiography and perfusion imaging of the hand. AJR Am J Roentgenol 2001;177: Planken RN, Tordoir JH, Dammers R, et al. Stenosis detection in forearm hemodialysis arteriovenous fistulae by multiphase contrast-enhanced magnetic resonance angiography: preliminary experience. J Magn Reson Imaging 2003;17: Planken RN, Tordoir JH, Kessels AG, et al. 3D CE-MR arteriography and selective venography of the upper extremity prior to hemodialysis access fistula surgery. In: Proceedings of the 12th Annual Meeting of ISMRM, Kyoto, Japan, 2004 (Abstract 2579). 19. Froger CL, Duijm LE, Liem YS, et al. Stenosis detection with MR angiography and digital subtraction angiography in dysfunctional hemodialysis access fistulas and grafts. Radiology 2005;234: Smits JH, Bos C, Elgersma OE, et al. Hemodialysis access imaging: comparison of flow-interrupted contrast-enhanced MR angiography and digital subtraction angiography. Radiology 2002;225: Born M, Willinek WA, Gieseke J, von Falkenhausen M, Schild H, Kuhl CK. Sensitivity encoding (SENSE) for contrast-enhanced 3D MR angiography of the abdominal arteries. J Magn Reson Imaging 2005;22: Hoogeveen RM, von Falkenhausen M, Gieseke J. Fast dynamic, high resolution contrast-enhanced MR angiography with CENTRA keyhole and SENSE. In: Proceedings of the 12th Annual Meeting of ISMRM, Kyoto, Japan, 2004 (Abstract 9). 23. Swan JS, Carroll TJ, Kennell TW, et al. Time-resolved three-dimensional contrast-enhanced MR angiography of the peripheral vessels. Radiology 2002;225: Mazaheri Y, Carroll TJ, Du J, et al. Combined time-resolved and high-spatial-resolution 3D MRA using an extended adaptive acquisition. J Magn Reson Imaging 2002;15: Kozerke S, Tsao J, Razavi R, Boesiger P. Accelerating cardiac cine 3D imaging using k-t BLAST. Magn Reson Med 2004;52:19 26.