JOURNAL OF MAGNETIC RESONANCE IMAGING 28:181–189 (2008)
Original Research
3D Nongadolinium-Enhanced ECG-Gated MRA of the Distal Lower Extremities: Preliminary Clinical Experience Ruth P. Lim, MBBS, MMed, FRANZCR,1* Elizabeth M. Hecht, MD,1 Jian Xu, BS,2 James S. Babb, PhD,1 Niels Oesingmann, PhD,2 Samson Wong, MD,1 Bart E. Muhs, MD, PhD,3 Paul Gagne, MD,3 and Vivian S. Lee, MD, PhD, MBA1 Purpose: To report our initial experience implementing a noncontrast-enhanced electrocardiograph (ECG) gated fast spin echo magnetic resonance angiography (MRA) technique for assessment of the calf arteries.
Conclusion: Our results indicate that when technically successful, noncontrast-enhanced MRA using ECG-gated fast spin echo can provide accurate imaging of the calf and pedal arteries. However, further development and optimization are needed to improve the robustness of the technique.
Materials and Methods: Noncontrast MRA images of 36 clinical patients examined over a 6-month period were evaluated by two radiologists for length and degree of stenosis of arterial segments. Diagnostic confidence in the technique was also recorded. The reference standard was a consensus reading by both radiologists using the noncontrast technique combined with two gadoliniumenhanced techniques: bolus-chase and time-resolved imaging.
Key Words: magnetic resonance angiography (MRA); noncontrast; peripheral arterial disease (PAD) J. Magn. Reson. Imaging 2008;28:181–189. © 2008 Wiley-Liss, Inc.
Results: For stenosis evaluation the noncontrast technique demonstrated accuracy 79.4% (1083/1364), sensitivity 85.4% (437/512), and specificity 75.8% (646/ 852). The sequence demonstrated high negative predictive value (92.3%, 646/700). The technique had serious artifacts leading to poor diagnostic confidence in 17 patients (47.2%). These included motion (n ⫽ 7) and artifacts specific to the sequence, including inaccurate trigger delays (n ⫽ 5), linear artifact (n ⫽ 7), and vessel blurring (n ⫽ 5). When only patients in whom there was satisfactory diagnostic confidence were considered, accuracy, sensitivity, and negative predictive value were 92.2% (661/717), 92.4% (158/171), and 97.5% (503/ 516), respectively.
1 Department of Radiology (MRI), New York University Medical Center, New York, New York. 2 Siemens Medical Solutions USA, Inc., MR R&D Collaboration, NYU CBI, New York, New York. 3 Department of Vascular Surgery, New York University Medical Center, New York, New York. Current address for B.E. Muhs: Yale Medical Group, Department of Vascular Surgery Yale Surgical Specialties, New Haven, Connecticut. *Address reprint requests to: R.L., Department of Radiology, 560 First Ave., HW-202, New York, NY 10016. E-mail:
[email protected] Received August 24, 2007; Accepted March 14, 2008. DOI 10.1002/jmri.21416 Published online in Wiley InterScience (www.interscience.wiley.com).
© 2008 Wiley-Liss, Inc.
PERIPHERAL ARTERIAL DISEASE (PAD) is a major cause of morbidity and mortality in the Western world, with an estimated prevalence up to 20% in those over 75 years of age (1). Accurate depiction of calf and pedal arteries is vital for planning revascularization procedures and to determine if amputation is indicated (2– 4). While x-ray angiography remains the gold standard for assessment of the peripheral arteries (5), contrast-enhanced magnetic resonance angiography (MRA) has emerged as a less invasive approach, with reported sensitivities and specificities of 84%–99.5% and 93.4%– 99%, respectively, for hemodynamically significant disease (6 –13) that has become the first-line modality for diagnosis of lower extremity peripheral vascular disease at most institutions. Recent reports have linked the gadolinium chelates used in contrast-enhanced MRI studies with the development of nephrogenic systemic fibrosis in patients with moderate or severe renal insufficiency or acute hepatorenal syndrome (14 –20). The frequency of renal insufficiency in the peripheral vascular disease population (21), coupled with preexisting concerns about cost and dose limitations of gadolinium contrast material in some countries (22), have made it imperative that accurate, reproducible, and time-efficient alternatives to gadolinium-enhanced MRA be developed and made widely available. Extending from the initial description of the approach by Wedeen et al (23), Miyazaki et al (24,25) developed a noncontrast-enhanced MRA technique that uses an
181
182
Lim et al.
Figure 1. MR arteriograms are achieved by subtracting systolic images from diastolic images. In diastole, slow arterial flow appears bright, whereas in systole there is arterial signal loss due to fast flowing blood (arrows). Venous signal is bright in both systole and diastole as there is slow flow throughout the cardiac cycle. Subtraction of the two datasets cancels out venous signal, leaving only arterial signal.
electrocardiograph (ECG) triggered three-dimensional partial Fourier acquisition fast spin echo sequence to exploit differential arterial and venous flow velocities during the cardiac cycle (24). We describe our experiences with the implementation and performance of this technique using a 3D half-Fourier single shot fast spin echo sequence in a clinical population with PAD. MATERIALS AND METHODS Patients All lower extremity MRAs in patients referred to our department for suspected or known PAD between September 2005 and February 2006 were retrospectively reviewed at least 2 months after the clinical interpretation. The study was HIPAA-compliant and a waiver of consent was approved by the Institutional Review Board. Of the 40 patients identified on review, four were excluded: two patients with tachyarrhythmias (heart rate greater than 100 beats per minute) and two patients imaged using incorrect phase-encoding directions. Of the final 36-patient cohort (23 male, 13 female, mean age 70.3 years, age range 37–97 years), clinical indications were: claudication (n ⫽ 24), critical ischemia (n ⫽ 7), suspected distal emboli (n ⫽ 2), cellulitis (n ⫽ 1), slowly healing ulcer (n ⫽ 1), and bilateral popliteal aneurysms (n ⫽ 1). MR Imaging Protocol All examinations were performed using a 1.5T MR system (Avanto or Symphony; Siemens, Erlangen, Germany) and a peripheral phased array 6-element coil anteriorly and laterally and spine coil posteriorly. All
patients underwent three MRA sequences of the infragenual arteries in the following order: nongadolinium-enhanced 3D MRA (NG), contrast-enhanced 3D time-resolved MRA, and bolus chase contrast-enhanced 3D MRA. A 500-mm field of view (FOV), covering tibial plateau to metatarsal bases, was applied with identical coronal and anterior-posterior coverage, slice thickness 0.8 – 1.0 mm, and parallel imaging (GRAPPA, acceleration factor 2–3) for all sequences. The oblique coronal imaging slab followed the expected anatomical plane of the infragenual arteries. NG images were acquired using an ECG-gated 3D half-Fourier fast spin echo sequence triggered for systolic and diastolic acquisitions. The readout direction was parallel to vessel orientation (head to foot) with central k-space reordering to enhance systolic flow dephasing (25). A readout flow-spoiling gradient of 25%–35% was also used to enhance the differences between arteries and veins, where this value represents a percentage of half the area of the readout gradient (25). Fast arterial flow appears dark and slow flow in diastolic-phase arteries and all veins appear bright. MR arteriograms were achieved by subtracting systolic from diastolic images using automated scanner-generated subtraction in the image domain (Fig. 1). Parameters used were: TR ⫽ 2 R-R intervals, effective TE ⫽ 49 –59 msec, refocusing flip angle ⫽ 101–180°, voxel size 0.8 –1.0 ⫻ 1.6 –2.0 ⫻ 2.1–2.6 mm, bandwidth 975 Hz/voxel, and acquisition matrix 256 –320 ⫻ 192–237, with a constant 500 ⫻ 500 mm FOV for each partition. The differing matrix sizes refer to the two different magnets used, Symphony (lower spatial resolution) and
Lower Extremity Noncontrast MRA
183
Table 1 Imaging Parameters of the Time-Resolved and Bolus Chase Contrast-Enhanced MRA Sequences Sequence
Time-Resolved MRA
Bolus Chase MRA
TR (msec) TE (msec) Flip angle (°) Matrix Partitions Voxel size (mm3) Bandwidth (Hz/pixel) Acquisition time (s) Number of measures View sharing (TREAT – Siemens Medical Solutions, Erlangen, Germany) (26)
2.5 1.0 15-25 512 x 208 52 interpolated to 104 0.8-1.0 x 1.0 x 1.5 890 9.4 8 (1 precontrast, 7 postcontrast) Yes, in n⫽27 (all patients performed on Avanto system, not available in 9 patients performed on Symphony system)
3.3 1.2 25-30 512 x 307 52 interpolated to 104 0.8-1.0 x 1.0 x 1.2 445 13.7 3 (1 precontrast, 2 postcontrast) No
Avanto (higher spatial resolution), but were fixed for a given subject for each of the sequences used. A nonselective inversion recovery pulse (inversion time [TI] ⫽ 170 msec) was used for background signal suppression. With 5-msec interecho spacing, the acquisition window for each partition using half-Fourier acquisitions in the phase-encoding direction was 325 msec. A total of 78 partitions were acquired and zero-filled to 104 partitions, for acquisition times of 2 minutes (range 1.5–3.0 min). Total acquisition time for a systolic and a diastolic acquisition was 4 minutes (3– 6 min). To determine trigger delays corresponding to systole and diastole, an inversion recovery 2D ECG-gated halfFourier fast spin echo sequence was repeated at 100msec trigger delay intervals. For systolic images we used an ECG trigger delay (TD) of 230 msec (including inversion time). The TD range for diastolic images was typically 630 –730 msec. The NG sequence was followed by time-resolved (26) and bolus chase (8) gadolinium-enhanced sequences, with parameters of the contrast-enhanced sequences outlined in Table 1. Time-resolved acquisitions were performed after injection of 10 mL (⬇0.05 mmol/kg patient body weight) of gadopentetate dimeglumine (Magnevist; Berlex Laboratories, Wayne, NJ) at 2 mL/ sec via automated injector (Spectris, Medrad, Indianola, PA) and a 20-mL saline flush. An empiric scan delay of the time to peak enhancement (measured using a 1-mL test dose) at the common femoral arteries minus 10 seconds was used to ensure no vascular enhancement in the first acquisition. Bolus chase MRA was performed last with three-station (abdominopelvic, thighs, and calves) moving table technology, with third station position and coverage identical to NG and timeresolved MRA. After time-resolved MRA, precontrast images were acquired at all stations followed by postcontrast bolus chase images after contrast injection of 20 mL at 2 mL/sec, then 10 mL at 1 mL/sec and then a 20-mL saline flush, with timing of postcontrast scans calculated from the prior timing examination. Image Evaluation Two radiologists retrospectively reviewed datasets in random order, blinded to patient identity and pathology
(EMH with 4 years of experience, and RPL, with 1 year of experience). Source, subtracted, and MIP images of all three sequences were viewed on a workstation (Syngo, Siemens, Germany). For NG, readers relied on subtracted images for interpretation. The sequence was assessed for diagnostic confidence on a 5-point scale in each patient (1 ⫽ uninterpretable, 2 ⫽ poorly identified with uncertain diagnosis, 3 ⫽ identified with probable diagnosis, 4 ⫽ clearly identified with highly probable diagnosis, 5 ⫽ definite diagnosis). For purposes of analysis, studies with diagnostic confidence of 1 or 2 were considered nondiagnostic. Factors hindering interpretation were recorded. The reference standard was the consensus opinion of both readers using all three MRA sequences, evaluated at least 2 weeks following individual sequence interpretations. The infragenual arteries were divided into 10 segments per leg: distal popliteal artery; tibioperoneal trunk; proximal and distal anterior tibial artery; proximal and distal peroneal artery; proximal and distal posterior tibial artery; dorsalis pedis artery; and plantar arteries. Each segment was assessed for greatest degree of stenosis using a grading system based on our standard clinical practice: 1 ⫽ 0%–29% stenosis (mild), 2 ⫽ 30%– 69% stenosis (moderate), 3 ⫽ 70%–99% stenosis (severe), and 4 ⫽ occlusion. Each segment was also graded for visualized length: 1 ⫽ 50%–100% of expected length of the segment, 2 ⫽ 1%– 49% of expected length, and 3 ⫽ occlusion of entire segment. Intraoperative angiographic correlation of the more symptomatic leg was available in five patients, with no intervening therapy or clinical change between studies. The time interval between MRA and x-ray angiography ranged from 19 – 67 days (mean 37 days). Angiographic images were retrospectively assessed by a vascular surgeon (BEM with 4 years of percutaneous interventional experience). Statistical Analysis The accuracy of stenosis and length assessments for the NG sequence was assessed relative to the reference standard; assessments were considered correct only if in exact agreement with the reference standard. Failure to visualize a segment on the NG sequence that was
184
Lim et al.
Factors that hindered visualization and decreased reader confidence in the NG sequence were: motion (n ⫽ 7), inaccurate trigger delays (n ⫽ 5) (Fig. 3), linear artifacts in the anterior to posterior phase encoding direction (n ⫽ 7), blurring (n ⫽ 5) (Fig. 4), and wrap artifact in the anterior to posterior phase-encoding direction (n ⫽ 4). Stenosis
Figure 2. MIP subtraction images of an 84-year-old male presenting with right leg claudication. a: Nongadolinium-enhanced image (2xRR/49/refocusing flip angle 104°). Small collateral vessels bridging an occlusion of the distal right peroneal artery are well depicted (arrow), as is bilateral disease of the anterior tibial arteries (arrowheads). b: Time-resolved image (2.5/1.0/15°) and (c) bolus chase image (3.3/1.2/25°) also demonstrate the peroneal artery occlusion (arrow) and anterior tibial artery disease (arrowheads).
present on the reference standard was considered an incorrect assessment for the sequence. Segments not included in the imaging FOV on any sequence imaging the calf station were excluded from analysis. For stenosis, a score of 3 or 4 (ⱖ70% stenosis) was designated positive and accuracy was characterized in terms of sensitivity, specificity, and positive and negative predictive values for the detection of positive stenosis. For length, accuracy was expressed as percentage concordance with the reference standard and in terms of estimation error, computed as the absolute value of the difference between assessments derived by the reference standard and the NG sequence. Cohen’s kappa was used to assess reader agreement. All reported Pvalues are two-sided and were declared significant when less than 0.05. SAS v. 9.0 (SAS Institute, Cary, NC) was used for all statistical computations.
Based on the reference standard interpretation, 256 of 682 segments had significant stenoses (37.5%). These were distributed relatively evenly distal to the tibioperoneal trunk. Using each of the two individual reader assessments for each of the 682 segments, 1083/1364 segments (79.4%) were correctly evaluated when the NG sequence was compared with the reference standard. NG sensitivity, specificity, positive and negative predictive values for stenoses ⱖ70% were 85.4% (437/512), 75.8% (646/852), 74.1% (437/590), and 92.3% (646/700), respectively. Segments considered nondiagnostic at NG were considered incorrect for the sequence for these calculations. The more detailed analysis below considers only those segments (1293) that were evaluated at NG by the readers. On a per-segment basis, overestimation of stenosis was most common in the proximal and distal peroneal and plantar artery segments. Fewest errors were recorded in the popliteal, proximal posterior tibial, and dorsalis pedis arteries. The results of segmental analysis are summarized in Table 2. For cases in which diagnostic confidence was greater than or equal to 3 (n ⫽ 19 patients, 52.8%), considered to be of diagnostic quality, a total of 717 of the 1293
RESULTS A total of 682 (94.7%) of 720 potential segments (20 per patient) were within the imaging FOV on review of all three sequences, comprising the reference standard. Figure 2 shows an example of NG compared with the contrast-enhanced sequences for a patient. Using individual reader assessments for each of the 682 segments, 1293 of a potential 1364 segments (94.8%) were evaluated with the NG sequence. The 71 segments not assessed using NG were predominantly at the most proximal and distal ends of the FOV (assessed as excluded from the FOV by readers), or when images were considered uninterpretable on a per-leg basis after reference to both subtraction and source datasets and no assessment of length or stenosis was made by the readers (Fig. 3). Diagnostic confidence for the sequence averaged 2.94 ⫾ 0.95 (mean ⫾ standard deviation [SD]) with a median of 3 on a 5-point scale. Seventeen cases (47.2%) had a diagnostic confidence of 1 or 2, while 19 were considered 3 or above, with the latter considered to be of diagnostic standard.
Figure 3. Single vessel runoff in a 90-year-old female presenting with bilateral claudication. a: Noncontrast-enhanced image clearly depicts runoff via the left peroneal artery, but the right leg images are nondiagnostic, due to faster inflow of this leg and intrusion of the diastolic acquisition onto the systolic phase of the second R-R interval (2xRR/49/refocusing flip angle 108°). b: Corresponding time-resolved contrast-enhanced image (2.5/1.0/15°) shows patency of the right popliteal artery, tibioperoneal trunk, and posterior tibial artery, with a focal proximal posterior tibial artery stenosis (arrow).
Lower Extremity Noncontrast MRA
185
Analysis of the 717 segments considered diagnostically interpretable (confidence ⬎2) demonstrated correct stenosis assessments in 573/717 segments (79.9%). The mean ⫾ SD of the error in stenosis assessments was 0.13 ⫾ 0.66 for this group, compared with 0.25 ⫾ 0.98 when all cases were included. For incorrect assessments, overestimation of stenosis was present in the majority (Table 2). The kappa coefficient showed substantial agreement (0.70) between readers for identification of significant stenosis using the NG sequence when all segments were evaluated. For the diagnostically interpretable group the kappa coefficient was 0.77. Length
Figure 4. Coronal MIP subtraction noncontrast-enhanced image of the left leg (2xRR/59/refocusing flip angle 180°) (a) in a 64-year-old male demonstrates vessel blurring in the phaseencoding direction that can occur due to flow attenuation characteristics on fast spin echo sequences. b: Comparable MIP subtraction time-resolved contrast-enhanced image (3.4/ 1.3/25°) demonstrating true vessel caliber.
segments were evaluated. The accuracy among these technically successful cases was high at 92.2% (661/ 717), with high sensitivity (92.4%, 158/171), specificity (92.1%, 503/546), and excellent negative predictive value (97.5%, 503/516). The positive predictive value was 78.6% (158/201) for this subgroup.
At the reference standard evaluation there was complete occlusion in 91 of 682 (13.3%) segments in our series, while in 59/682 (8.7%), 1%– 49% of expected segment length was present. NG demonstrated correct length assessments in 986/ 1364 segments (72.3%), with a mean and SD of the absolute error of length assessment of 0.421 ⫾ 0.74. When compared with the reference standard, visualized segmental length was predominantly underestimated by the NG sequence, that is, higher scores for length were recorded for NG. When segments that were assessable on the NG sequence were evaluated there was underestimation of length using the NG sequence for 271 out of 1293 segments (21.0%). There was overestimation of length using the NG sequence in 30 of the 1293 segments (2.3%). Errors were most common in proximal and distal peroneal artery and plantar artery segments, summarized in Table 3. Less error was observed when patients with high diagnostic confidence were compared with those in whom diagnostic confidence was low. In the subgroup in whom diagnostic confidence was greater than 2 (n ⫽ 19), correct length assessments were observed in 87.3% (626/717) with a mean and SD of the absolute error of length assessment of 0.173 ⫾ 0.48.
Table 2 Segmental Analysis of Over- and Underestimations of Stenosis Compared With the Reference Standard Diagnostic Confidence High (Confidence ⫽ 3-5)
Popliteal Tibioperoneal trunk Proximal posterior tibial artery Distal posterior tibial artery Proximal peroneal artery Distal peroneal artery Proximal anterior tibial artery Distal anterior tibial artery Dorsalis pedis artery Plantar artery Total
All Cases
Number of Overestimated Segments
Number of Underestimated Segments
Number of Overestimated Segments
Number of Underestimated Segments
7.0% (4/57) 15.1% (11/73) 17.1% (13/76) 10.5% (8/76) 15.8% (12/76) 21.1% (16/76) 16.2% (11/68) 10.5% (8/76) 8.7% (6/69) 15.7% (11/70) 13.9% (100/717)
7.0% (4/57) 8.2% (6/73) 6.6% (5/76) 5.3% (4/76) 7.9% (6/76) 3.9% (3/76) 11.8% (8/68) 3.9% (3/76) 2.9% (2/69) 4.3% (3/70) 6.1% (44/717)
18.5% (17/92) 20.6% (26/126) 18.3% (26/142) 19.0% (27/142) 26.1% (37/142) 28.9% (41/142) 19.5% (25/128) 23.9% (34/142) 18.5% (20/108) 26.4% (34/129) 22.2% (287/1293)
8.7% (8/92) 9.5% (12/126) 12.7% (18/142) 4.9% (7/142) 12.0% (17/142) 8.5% (12/142) 10.9% (14/128) 7.7% (11/142) 7.4% (8/108) 4.7% (6/129) 8.7% (113/1293)
186
Lim et al.
Table 3 Segmental Analysis of Underestimation and Overestimation of Visualized Length Compared With the Reference Standard Diagnostic Confidence High (Confidence ⫽ 3-5)
Popliteal Tibioperoneal trunk Proximal posterior tibial artery Distal posterior tibial artery Proximal peroneal artery Distal peroneal artery Proximal anterior tibial artery Distal anterior tibial artery Dorsalis pedis artery Plantar artery Total
All Cases
Number of Underestimated Segments
Number of Overestimated Segments
Number of Underestimated Segments
Number of Overestimated Segments
5.3% (3/57) 5.5% (4/73) 7.9% (6/76) 6.6% (5/76) 19.7% (15/76) 23.7% (18/76) 8.8% (6/68) 7.9% (6/76) 7.2% (5/69) 24.3% (17/70) 11.9% (85/717)
0% (0/57) 0% (0/73) 1.3% (1/76) 1.3% (1/76) 0% (0/76) 0% (0/76) 1.5% (1/68) 0% (0/76) 1.4% (1/69) 2.9% (2/70) 0.8% (6/717)
15.2% (14/92) 10.3% (13/126) 12.0% (17/142) 19.0% (27/142) 28.9% (41/142) 33.1% (47/142) 16.4% (21/128) 21.1% (30/142) 19.4% (21/108) 31.0% (40/129) 21.0% (271/1293)
0% (0/92) 1.6% (2/126) 5.6% (8/142) 2.8% (4/142) 2.1% (3/142) 2.1% (3/142) 1.6% (2/128) 1.4% (2/142) 2.8% (3/108) 2.3% (3/129) 2.3% (30/1293)
The kappa coefficient showed substantial agreement between readers for the NG sequence in terms of length (0.65) assessments, with a higher kappa coefficient when only diagnostically interpretable patients were considered (0.70). X-Ray Angiography Correlation Single leg correlation was available in five patients (Figs. 5, 6). A total of 46 of 50 potential segments were visible at x-ray angiography, with three excluded from the imaging FOV and one DP segment within the imaging FOV by MR but not visualized at angiography. All 50 were present at reference standard MRA. For stenosis assessment there was disagreement in 3 of 46 segments (6.5%), where reference standard MRA overestimated stenosis by 1 grade compared to angiography in 2 segments (proximal posterior tibial and peroneal arteries), and by 2 grades in 1 distal peroneal artery segment. There was good agreement between the MRA reference standard and angiography regarding segment length (42 out of 46 segments, 91.3%). In three of the discrepant segments, two distal peroneal and one distal anterior tibial artery segment, a greater length of the vessel was appreciated at MRA compared with angiography (Fig. 6). For NG, only 44 out of 46 segments were compared with angiography, as two segments were not included in the NG FOV, and of these, 29 of 44 segments (65.9%) were concordant with angiography for stenosis assessment. If patients with diagnostic quality studies only were considered, agreement between NG and angiography was improved to 14/17 (82.4%). For length assessment, there was agreement in 27 of 44 segments (61.4%), which improved to 15/17 (88.2%) if only diagnostic NG studies were considered. MRA correlation with angiography is summarized in Table 4. DISCUSSION Nongadolinium-enhanced peripheral MRA methods have been explored and utilized in the past. Two-dimensional time-of-flight MRA was used before gadolin-
Figure 5. Images of the left leg of a 74-year-old female presenting with left leg claudication with grayscale inverted for closer correlation with the angiographic image. There is two-vessel runoff. a: MIP subtraction noncontrast-enhanced image, obtained with patient heart rate of 90 beats per minute (2xRR/51/refocusing flip angle 108°) clearly demonstrates two-vessel runoff via the anterior tibial (arrow) and posterior tibial (arrowhead) arteries, with an occluded peroneal artery, with patency of the plantar artery and reconstituted dorsalis pedis artery clearly apparent, sufficient information on which to proceed with further intervention. b: Intraoperative angiogram image confirming the MRA findings performed prior to balloon angioplasty of a severe left external iliac artery stenosis.
Lower Extremity Noncontrast MRA
187
Figure 6. An 85-year-old male presenting with a nonhealing left foot ulcer 2 years after left femoropopliteal bypass graft. a: MIP subtraction bolus chase contrast-enhanced image (3.3/1.2/25°) demonstrates single vessel runoff on the left via the posterior tibial artery, with a high-grade stenosis of the popliteal artery just distal to the distal graft anastomosis (arrow), and a short segment tibioperoneal trunk occlusion (arrowhead). The proximal peroneal and anterior tibial arteries are occluded, with reconstitution of very narrow caliber distal peroneal and anterior tibial artery segments via collateral flow. b: MIP subtraction noncontrast-enhanced image (2xRR/49/refocusing flip angle 108°) demonstrates the popliteal artery stenosis (arrow) and tibioperoneal trunk occlusion (arrowhead). The plantar artery and reconstituted distal peroneal and posterior tibial segments are not seen; however, sufficient detail is present such that the same management recommendations would have been made. In this case the patient was recommended for and underwent percutaneous intervention. c: Composite intraoperative angiogram image confirming the presence of popliteal artery stenosis (arrow) and tibioperoneal trunk occlusion (black arrowhead). Note that the reconstituted distal peroneal and anterior tibial segments are better seen on the bolus chase MRA. d: Digital subtraction angiography image following balloon angioplasty of the popliteal artery stenosis (arrow) with a good technical and clinical result.
ium-enhanced MRA was established as the standard MRA technique, with good results (27). However, the technique is hampered by long imaging times and sensitivity to flow turbulence and in-plane signal saturation (28). Similarly, phase contrast peripheral MRA (29) also suffers from relatively long imaging times. Steadystate free precession (SSFP) techniques have been described in imaging of other vascular beds (30,31), but have not been reported for peripheral MRA to our knowledge. The infragenual region in particular offers the challenge of closely apposed, small caliber arteries
and veins, which are difficult to separate using basic SSFP techniques. A half-Fourier fast spin echo approach, ECG-gated, noncontrast MRA was initially described in the chest and abdomen by Miyazaki and colleagues (22,24), and more recently in the peripheral arteries of healthy volunteers (25). It has shorter imaging time than time-offlight and phase-contrast techniques over a large FOV, and is able to separate arteries from veins, and therefore may present a valuable noncontrast peripheral MRA alternative. To our knowledge, clinical assessment
Table 4 Agreement of MRA Sequences With X-Ray Angiography (XRA) for Stenosis and Length Assessments, and Segments Where There Was Disagreement With Angiography
Sequence
All NG Diagnostic NG (confidence ⱖ 2) MRA reference standard
Agreement With XRA (Length)
Underestimation of Vessel Length Compared With XRA (Segments)
Overestimation of Vessel Length Compared With XRA (Segments)
4
27/44 (61.4%)
16
1
2
1
15/17 (88.2%)
2
0
3
0
42/46 (91.3%)
1
3
Agreement With XRA (Stenosis)
Overestimation of Stenoses Compared With XRA (Segments)
Underestimation of Stenoses Compared With XRA (Segments)
29/44 (65.9%)
11
14/17 (82.4%)
43/46 (93.5%)
NG, nongadolinium-enhanced MRA.
188
of this method in a symptomatic population has not been published. We calculated an accuracy of 79.4% (1083/1364) for stenosis in our series, which improved to 92.2% (661/717) when only cases considered to be of diagnostic quality (n ⫽ 19, 52.8%) were included. Overestimation of stenosis and underestimation of length were the main errors demonstrated in our series. The majority of stenosis overestimations were observed in the peroneal and plantar artery segments. The majority of length underestimations were again demonstrated in the peroneal and plantar artery segments. We postulate that the small caliber of these arteries could contribute to the greater degree of error observed. Our analysis suggests that the most important cause of error was imperfect timing of the triggered acquisitions. Inappropriate selection of the systolic trigger delay leads to suboptimal flow voids on systolic images, and the resulting overestimation of stenosis and underestimation of vessel length, on subtracted datasets. This problem was likely exacerbated by our reliance on subtraction images for evaluation of NG images. While review of source images could theoretically have helped to overcome this limitation, in practice, interpretation of nonsubtracted images is often limited because of the close proximity of small caliber arteries and veins below the knee. Selection of an inappropriate diastolic TD also led to suboptimal images in five patients in our series. Selection of trigger delays was based on the use of a 2D ECG-triggered sequence that sampled trigger delays at intervals of 100 msec and incorporated an inversion pulse for fat suppression. We have subsequently replaced the trigger delay scout with a through-plane phase contrast sequence to obtain an arterial velocity curve that enables identification of optimal systolic and diastolic trigger delays within the cardiac cycle. Two other technical problems that can be addressed in the future further contributed to a lack of robustness with this technique. The flow-spoiling gradient in the craniocaudal direction was typically fixed to 25%–35%. Miyazaki et al (25) have shown that improved visualization of vessels with varying flow can be achieved by customized selection of flow-spoiling gradients to match specific vascular territories in different patients. Additionally, the relatively long sampling windows during each cardiac cycle (325 msec per partition in our series) resulted in degraded image quality, with motion artifacts (n ⫽ 7) and blurring (n ⫽ 5) commonly seen. Higher parallel imaging factors, particularly at higher field strengths, may help overcome these problems by reducing the number of echoes per partition and consequently reduce vessel blurring by reducing T2-blurring effects and minimizing motion-related artifacts (32). Patients with arrhythmias were also problematic. With tachyarrhythmias, the sampling window of 325 msec per partition occupied an increasing proportion of the RR interval, with subsequent poor temporal separation of putative systolic and diastolic acquisitions. Shorter acquisition windows could potentially make this sequence less sensitive to tachyarrhythmias, allowing for more precise acquisition during peak systolic and slowest diastolic flow. Irregular rhythms, such as
Lim et al.
atrial fibrillation, would provide an added challenge, as the technique relies on prescription of fixed trigger delays. Also noted was linear artifact in the phase-encoding direction seen only in the nine patients scanned with one of our systems (Symphony). Further testing with phantoms showed that this was an inherent problem with the sequence on that system, with some contribution from the use of parallel imaging, potentially related to eddy currents during gradient switching. Although in the current implementation the method tends to overestimate disease, our results do show a high negative predictive value (646/700 ⫽ 92.3%) with NG that suggests that it could have a potential role in screening patients. In the event of a negative examination, the NG method could obviate progression to a contrast examination including angiography. Separate from the problems observed with the sequence, our study has recognized limitations, including a lack of angiographic correlation for most patients. At our institution, MRA findings often dictate management, with angiography performed only if intervention is being considered, creating an inherent selection bias. Angiography, however, is an imperfect gold standard for the calves, as patent distal vessels are not always wellopacified, as was shown with the angiographic studies available for comparison in our series. Also, the system used to grade stenosis was modeled on our clinical practice, and stenoses in the 50%– 69% range were not included as positive findings, although these are often considered hemodynamically significant in the literature. The order in which sequences were performed could not be randomized in our study, as gadolinium’s T2 shortening effect would degrade NG images. Finally, we evaluated only the accuracy of nongadolinium-enhanced MRA in the infragenual station. Evaluation of three-station imaging in clinical patients using the technique is the subject of future work. In conclusion, in our initial evaluation the noncontrast-enhanced ECG-gated 3D half-Fourier fast spin echo MRA sequence shows high negative predictive value for evaluation of infragenual arteries. When technically successful, the approach provides results that are reasonably comparable to contrast-enhanced techniques, even in the setting of severe disease, and may be sufficient for accurate clinical decision-making. In its current implementation, we found the method to be technically successful in just over half of all cases. With refinement, such as optimized trigger delay selection, shortened acquisition window, and tailored flow-spoiling gradients, we believe the method’s technical robustness can be substantially improved.
ACKNOWLEDGMENTS We thank Dr. Mitsue Miyazaki of the MR Engineering Department, Medical Systems Division, Toshiba, Tochigi, Japan, for helpful advice regarding implementation and suggestions for improvement of the noncontrast half-Fourier fast spin echo sequence.
Lower Extremity Noncontrast MRA
REFERENCES 1. Pasternak RC, Criqui MH, Benjamin EJ, et al. Atherosclerotic Vascular Disease Conference: Writing Group I: epidemiology. Circulation 2004;109:2605–2612. 2. Hughes K, Domenig CM, Hamdan AD, et al. Bypass to plantar and tarsal arteries: an acceptable approach to limb salvage. J Vasc Surg 2004;40:1149 –1157. 3. Aulivola B, Pomposelli FB. Dorsalis pedis, tarsal and plantar artery bypass. J Cardiovasc Surg (Torino) 2004;45:203–212. 4. Pomposelli FB, Kansal N, Hamdan AD, et al. A decade of experience with dorsalis pedis artery bypass: analysis of outcome in more than 1000 cases. J Vasc Surg 2003;37:307–315. 5. Borrello JA. MR angiography versus conventional X-ray angiography in the lower extremities: everyone wins. Radiology 1993;187: 615– 617. 6. Baum RA, Rutter CM, Sunshine JH, et al. Multicenter trial to evaluate vascular magnetic resonance angiography of the lower extremity. American College of Radiology Rapid Technology Assessment Group. JAMA 1995;274:875– 880. 7. Ruehm SG, Hany TF, Pfammatter T, Schneider E, Ladd M, Debatin JF. Pelvic and lower extremity arterial imaging: diagnostic performance of three-dimensional contrast-enhanced MR angiography. AJR Am J Roentgenol 2000;174:1127–1135. 8. Meaney JF, Ridgway JP, Chakraverty S, et al. Stepping-table gadolinium-enhanced digital subtraction MR angiography of the aorta and lower extremity arteries: preliminary experience. Radiology 1999;211:59 – 67. 9. Steffens JC, Schafer FK, Oberscheid B, et al. Bolus-chasing contrast-enhanced 3D MRA of the lower extremity. Comparison with intraarterial DSA. Acta Radiol 2003;44:185–192. 10. Bezooijen R, van den Bosch HC, Tielbeek AV, et al. Peripheral arterial disease: sensitivity-encoded multiposition MR angiography compared with intraarterial angiography and conventional multiposition MR angiography. Radiology 2004;231:263–271. 11. Leiner T, Kessels AG, Nelemans PJ, et al. Peripheral arterial disease: comparison of color duplex US and contrast-enhanced MR angiography for diagnosis. Radiology 2005;235:699 –708. 12. Vavrik J, Rohrmoser GM, Madani B, Ersek M, Tscholakoff D, Bucek RA. Comparison of MR angiography versus digital subtraction angiography as a basis for planning treatment of lower limb occlusive disease. J Endovasc Ther 2004;11:294 –301. 13. Loewe C, Schoder M, Rand T, et al. Peripheral vascular occlusive disease: evaluation with contrast-enhanced moving-bed MR angiography versus digital subtraction angiography in 106 patients. AJR Am J Roentgenol 2002;179:1013–1021. 14. Grobner T. Gadolinium—a specific trigger for the development of nephrogenic fibrosing dermopathy and nephrogenic systemic fibrosis? Nephrol Dial Transplant 2006;21:1104 –1108. 15. Gadolinium-containing contrast agents for magnetic resonance imaging (MRI): Omniscan, OptiMARK, Magnevist, ProHance and MultiHance. In: United States Food and Drug Administration (FDA): Public Health Advisory, 2006. 16. Sadowski EA, Bennett LK, Chan MR, et al. Nephrogenic systemic fibrosis: risk factors and incidence estimation. Radiology 2007; 243:148 –157.
189 17. Broome DR, Girguis MS, Baron PW, Cottrell AC, Kjellin I, Kirk GA. Gadodiamide-associated nephrogenic systemic fibrosis: why radiologists should be concerned. AJR Am J Roentgenol 2007;188: 586 –592. 18. Perazella MA, Rodby RA. Gadolinium use in patients with kidney disease: a cause for concern. Semin Dial 2007;20:179 –185. 19. Pedersen M. Safety update on the possible causal relationship between gadolinium-containing MRI agents and nephrogenic systemic fibrosis. J Magn Reson Imaging 2007;25:881– 883. 20. Marckmann P, Skov L, Rossen K, Heaf JG, Thomsen HS. Casecontrol study of gadodiamide-related nephrogenic systemic fibrosis. Nephrol Dial Transplant 2007;22:3174 –3178. 21. O’Hare AM, Bertenthal D, Shlipak MG, Sen S, Chren MM. Impact of renal insufficiency on mortality in advanced lower extremity peripheral arterial disease. J Am Soc Nephrol 2005;16:514 –519. 22. Urata J, Miyazaki M, Wada H, Nakaura T, Yamashita Y, Takahashi M. Clinical evaluation of aortic diseases using nonenhanced MRA with ECG-triggered 3D half-Fourier FSE. J Magn Reson Imaging 2001;14:113–119. 23. Wedeen VJ, Meuli RA, Edelman RR, et al. Projective imaging of pulsatile flow with magnetic resonance. Science 1985;230:946 – 948. 24. Miyazaki M, Sugiura S, Tateishi F, Wada H, Kassai Y, Abe H. Non-contrast-enhanced MR angiography using 3D ECG-synchronized half-Fourier fast spin echo. J Magn Reson Imaging 2000;12: 776 –783. 25. Miyazaki M, Takai H, Sugiura S, Wada H, Kuwahara R, Urata J. Peripheral MR angiography: separation of arteries from veins with flow-spoiled gradient pulses in electrocardiography-triggered three-dimensional half-Fourier fast spin-echo imaging. Radiology 2003;227:890 – 896. 26. Fink C, Ley S, Kroeker R, Requardt M, Kauczor HU, Bock M. Time-resolved contrast-enhanced three-dimensional magnetic resonance angiography of the chest: combination of parallel imaging with view sharing (TREAT). Invest Radiol 2005;40:40 – 48. 27. Yucel EK, Kaufman JA, Geller SC, Waltman AC. Atherosclerotic occlusive disease of the lower extremity: prospective evaluation with two-dimensional time-of-flight MR angiography. Radiology 1993;187:637– 641. 28. Kaufman JA, McCarter D, Geller SC, Waltman AC. Two-dimensional time-of-flight MR angiography of the lower extremities: artifacts and pitfalls. AJR Am J Roentgenol 1998;171:129 –135. 29. Steffens JC, Link J, Muller-Hulsbeck S, Freund M, Brinkmann G, Heller M. Cardiac-gated two-dimensional phase-contrast MR angiography of lower extremity occlusive disease. AJR Am J Roentgenol 1997;169:749 –754. 30. Deshpande VS, Shea SM, Laub G, Simonetti OP, Finn JP, Li D. 3D magnetization-prepared true-FISP: a new technique for imaging coronary arteries. Magn Reson Med 2001;46:494 –502. 31. Coenegrachts KL, Hoogeveen RM, Vaninbroukx JA, et al. Highspatial-resolution 3D balanced turbo field-echo technique for MR angiography of the renal arteries: initial experience. Radiology 2004;231:237–242. 32. Xu J, Oesingmann N, Stemmer A, et al. Reduced acquisition window with parallel technique improves non contrast 3D HASTE MRA imaging. In: Proc 14th Meeting ISMRM, Seattle, 2006; Abstract 1931.
