A prospective study using contrast enhanced turbo-FLASH (Fast Low-Angle Shot) magnetic resonance. (MR) angiography was performed to assess the arterial ...
1993, The British Journal of Radiology, 66, 1103-1110
Fast magnetic resonance angiography using turboFLASH sequences in advanced aortoiliac disease 1 2
U M SIVANANTHAN, MRCP, FRCR, 2 J P RIDGWAY, PhD, 2K BANN, DCR, 1S P VERMA, MD, MRCP, J CULLINGWORTH, HDCR, 2 J WARD, MSc, DCR and 1 M R REES, MRCP, FRCR
department of Cardiac Radiology and Cardiac Research Unit, Killingbeck Hospital, York Road, Leeds LS14 6UQ and 2 MRI Unit, St James's University Hospital, Leeds, UK Abstract
A prospective study using contrast enhanced turbo-FLASH (Fast Low-Angle Shot) magnetic resonance (MR) angiography was performed to assess the arterial anatomy in patients who had advanced atherosclerotic aortoiliac disease. This new imaging sequence was employed in 17 patients and the results were compared with conventional abdominal aortography. MR angiography accurately detected all aortic occlusions (3/3), their sites and their extent. All nine iliac occlusions were correctly identified (sensitivity 100%, specificity of 90%). The sensitivity was 100% for stenosis of 50% or greater in the abdominal aorta, and the iliac and common femoral arteries. The degree of stenosis was overgraded in 20 of 51 lesions (39.2%). Mild stenosis was overgraded as moderate stenosis in 10 and as severe stenosis in three. Moderate stenosis was overgraded as severe stenosis in four. None of the mild or moderate stenoses resulted in areas of signal voids suggestive of occlusions. Three severe stenoses were seen as areas of signal voids (two iliac, one femoral). In the eight patients who had in total 10 aneurysmal dilatations of the aorta or the iliac arteries, MR angiography was superior in demonstrating the true extent of the aneurysms. We conclude that turbo-FLASH MR angiography has the potential to be a useful non-invasive imaging technique for patients with advanced aortoiliac disease.
Invasive angiographic techniques have remained the definitive diagnostic and screening methods for peripheral vascular disease despite the increased use of ultrasound. There remains a need for a reliable noninvasive diagnostic technique as the demand for angiographic facilities increases to allow for the rapid development of angioplasty. The potential of MRI as a non-invasive technique for the evaluation of cardiovascular disease has been known for some time but only recently has been developed to the extent that it is clinically useful. MR angiography is now capable of depicting the vascular system in images similar to those of X-ray angiography. Numerous techniques have been developed to maximize signal from flowing blood while minimizing the signal from stationary tissue. These techniques have been successfully applied in the head and neck and are being developed for other areas in the vascular system. One major disadvantage of current MR angiographic techniques is that they are time-consuming. The advantages of application of fast MRI imaging techniques to the vascular system include elimination of motion artifacts, increased throughput, and patient comfort. Moreover, fast imaging provides the temporal resolution crucial for
Received 21 May 1993 and accepted 2 July 1993. Address correspondence to Dr M R Rees. Vol. 66, No. 792
dynamic studies. Preliminary studies suggest that fast MR angiography may have a useful role in atherosclerotic aortoiliac disease. Conventional MR angiography is not well suited to imaging of the abdominal aorta and iliac arteries because of the need to image a large volume of tissue, the prominence of fat in many patients, the visceral motion from respiration and peristalsis, and the extreme pulsatility of arterial flow in this region (Wendt et al, 1990). The turbo-FLASH sequence (Siemens Medical Engineering, Erlangen), based on the Snapshot FLASH pulse sequence (Haase, 1990), enables the acquisition of an image in as little as a fraction of a second by using repetition times of around 10 ms and typical echo times (TEs) of between 2-4 ms. Alternatively, if used with a multiple-slice acquisition mode, the very short TE of these sequences allows the acquisition of a greater number of slices, within a given breath-hold period, compared with a FLASH sequence using the same repetition time and acquisition matrix. We have used a multiple-slice turbo-FLASH pulse sequence to perform dynamic, breath-hold imaging of the aorta and iliac vessels following intravenous injection of an MRI contrast agent. We present our preliminary data on the combination of using a turbo-FLASH imaging sequence combined with an MRI contrast agent (gadolinium-DPT A) to enhance the blood signal in the diagnosis of aorto-iliac atherosclerotic disease. 1103
UM Sivananthan, J P Ridgway, K Bann et al
Materials and methods Patient selection We have employed 2-D Turbo-FLASH gradient echo sequences with gadolinium diethylene-triamine pentaacetic acid (Gd-DPTA) enhancement to image advanced aortoiliac disease in 17 patients. All patients selected had previously undergone abdominal aortography by means of conventional angiography within the prior 4 weeks. All patients had consented to the MR study. The standard contraindications for MR {e.g. pacemaker) were the main exclusion criteria for the study. Patients who had metallic vascular endoprosthesis, i.e. stents, inserted during the past 3 months were also excluded from the study. The patient population consisted of 15 men and two women, with a mean age of 64 years (range 56 to 72). One individual had a Palmaz iliac stent which was placed 4 months previously.
Figure 1. (a) Spin echo localizer showing the graphical positioning of 11 slices used during the dynamic series to include the full extent of the aorta iliac vessels, (b) 11 slices from the first acquisition of a dynamic series obtained in a patient with a Y-graft. Segments of the aorta and graft can be seen on most slices with high contrast. 1104
Imaging methods The technique was implemented on a Siemens MAGNETOM 42 SP MRI system operating at a field strength of 1.0 Tesla. Images were acquired using the system body coil for transmission and reception. A sagittal localizer was initially obtained to determine the best coronal or coronal/oblique plane of the aorta (Figure la). All the subsequent images were then obtained in this plane using a turbo-FLASH sequence (repetition time (TR) = 100 ms; echo time (TE) = 4 ms; flip angle = 80°; bandwidth = 355 Hz/pixel with asymmetric sampling). The minimum available slice thickness and field of view for this sequence is 4 mm and 350 mm, respectively. For the aorto-iliac studies, eleven 5 mm-thick slices with a 0.5 mm gap between them were acquired simultaneously using a 160x256 acquisition matrix and with the patient breath-holding for a period of 19 s. A 400 mm field of view was used which extended from the diaphragm to the inguinal ligaments. A preliminary unenhanced acquisition was performed to confirm correct positioning of the slices. We administered 0.1 mmol kg" 1 body weight of gadolinium diethylene triamine penta-acetic acid (Gd-DPTA) (MAGNEVIST, Schering Health Care) as an intravenous bolus via an antecubital vein, followed immediately by a saline flush. The period of the contrast injection was approximately 5 s. The first acquisition of the dynamic series was started 10 s after the end of injection of the contrast medium. We obtained four acquisitions with a 10 s inter-scan delay to allow the patient to breathe. All scanning was performed on arrested expiration as this is a more reproducible phase. The resultant images could be viewed as individual slices (Figure lb) and were also post-processed using a maximum intensity projection (MIP) algorithm to produce projection images (Figure 2) similar to the conventional angiograms (Laub, 1990). The MIP algorithm uses a ray tracing technique to determine the maximum intensity through the block of slices when viewed from a particular angle. This algorithm assumes that the highest intensity pixels will belong to the vessels The British Journal of Radiology, December 1993
Fast MR angiography using turbo-FLASH sequences
Figure 2. Maximum intensity projections produced from (a) seven selected slices of the first acquisition, and (b) five selected slices of the third acquisition. Venous return can be seen on the third acquisition although the arterial signal intensity is significantly reduced by dilution of the contrast medium. Smaller vessels such as the renal arteries are masked on the third acquisition by the background signal intensity.
containing contrast medium. With this method, the set of images is viewed along a series of user-defined projection angles, which include conventional, as well as selected, projections such as lateral views. Satisfactory results were achieved in all our patients (Figure 3). The total imaging time per patient was less than 10 min. Conventional abdominal aortography was performed with 70 ml of non-ionic iodinated contrast media delivered at a rate of 10 ml s~', via a 5-F pigtail catheter positioned at the level of the renal arteries. Data analysis The individual unprocessed two dimensional sections and the MIP images were correlated with the anteroposterior abdominal aortograms, all of which were available in all patients. The MR angiograms and the aortograms were analysed separately by two independent radiologists and the results were correlated. The most normal looking segments of aorta, iliac arteries and the common femoral arteries were measured on the conventional X-ray angiograms using the "pig-tail" catheter as a scaling device and the results were corrected for magnification accordingly. The corresponding segments on the projected images of the MR angiograms were also measured on the screen using distance measurement software. The results were analysed in three separate segments; aorta, iliac and common femoral. Four grades were used to categorize the atherosclerotic lesions with luminal compromise, i.e. stenosis, based on the most Vol. 66, No. 792
severe reduction of arterial diameter when compared with the most normal appearing segment immediately proximal or distal to the lesion. The grades were (a) mild stenosis (less than 50%), (b) moderate stenosis (5074%), (c) severe stenosis (75-99%) and (d) occlusion. This grading was employed for lesions in the aorta, iliac arteries and common femoral arteries. The length of the stenoses and the occlusions were tabulated separately for MR angiograms and for aortograms. The presence of a focal or diffuse signal void and other artifacts on the MR images were also documented. Results
Diameter of undiseased segments In all 17 patients the corrected diameters of the normal or the near normal looking segments of the aorta on conventional angiograms were analysed with the measured diameters of the corresponding MR angiograms. The results showed on average an 8.35 + 5.02% (range 0-25%) reduction in the diameter. In 15 patients who had patent iliac vessels, the diameters of the normal iliac segments were analysed. The results showed that the projected images of the MR angiograms underestimated the diameters by, on average, 14.67 + 8.96% (range 0-40%) in the common iliac segments, and by 18.8+12.37% (range 0-50%) in the external iliac segments. In two patients with aortic occlusion there was no significant reconstitution of the iliac arteries. The diameter of the 20 normal common 1105
UM Sivananthan, J P Ridgway, K Bonn et al
(a)
(b)
Figure 3. Atheromatous aortoiliac disease as seen on conventional aortogram (a) and on the magnetic resonance angiography using the maximum intensity projection display (b).
femoral segments were analysed in 10 patients, the MR angiograms showed on average a 29+15.53% (range 10-60%) reduction in diameter compared with the corrected values of the conventional angiograms. In the other seven patients; in two the common femoral arteries were outside the scanning planes on the MR images, in three there was significant degradation of the image quality owing to artifacts on the projected images, and in the other two the vessels were occluded. Aortic lesions Three patients were shown to have infra-renal aortic occlusions by conventional angiography, the site and extent of occlusions and the sites of reconstitution (two external iliac, one common iliac) were all correctly identified on MR angiography (Figure 4). In addition to this, in two patients thrombosed atherosclerotic abdominal aortic aneurysms were identified as the cause of the occlusion (Figure 5). In the other the full extent of the occluded normal diameter aorta could be seen in the pre-contrast images. Five other patients had aneurysmal dilatation of the lower end of the aorta, in all not only the luminal diameter but also the outer dimensions 1106
could be well seen. There were 13 stenotic lesions seen on conventional angiography, 11 were correctly graded (84.6%) by MR. In lesions of > 50% stenosis there were two false positives (mild lesions graded as moderate) but no false negatives. Overall MR angiography had a sensitivity of 100% and a specificity of 87.5%. MR angiography was superior in evaluating the aortic bifurcation, in three patients there was significant luminal compromise owing to indentation by posterior atheromatous plaques. Iliac lesions Sites of all nine iliac occlusions (seven common iliac, two external iliac) were correctly identified although there was slight overestimation of their lengths 4.56 + 2.88% (range 0-8%). All three common iliac aneurysms were well demonstrated on MR images (Figure 6). 19 of the 35 stenoses were correctly graded. There were no false negatives. 16 lesions were overgraded, seven mild lesions as moderate, three mild lesions as severe, and four moderate stenoses as severe. Two severe stenoses resulted in areas of signal voids suggestive of occlusions. If lesions of 50% or more The British Journal of Radiology, December 1993
Fast MR angiography using turbo-FLASH sequences
(a) (b) Figure 4. Conventional aortogram performed via the brachial route showing the infra-renal aortic occlusion (a), and the MR angiogram in the same patient (b).
stenosis is taken, MR angiography had a sensitivity of 100% and a specificity of 70%. Common femoral lesions There were two common femoral artery occlusions, both were correctly identified by MR angiography, but the areas of reformations were not seen as they were outside the field of view. The MIP images and the conventional angiograms were compared in the other 15 patients. Comparable images were present in 10, the common femoral vessels were outside the scanning planes in two and in three patients there was significant artifactual degradation of these areas. Of the three stenoses compared one was graded correctly, one mild stenosis was graded as moderate and one severe stenosis led to an area of signal void. Renal arteries Although the study was not designed to evaluate the renal arteries, they were clearly seen in the majority of patients either partially or throughout their full length. The number of renal arteries were correctly identified in all 11 patients who had renal areas included in their Vol. 66, No. 792
aortograms. Of the 31 renal arteries seen on the MR angiograms, 10 arteries could be seen throughout their full length, and 21 were seen partially. Renal outlines were well delineated in all 17 patients on the MR images, whereas in 11 conventional angiograms, the renal outlines were included in the abdominal films. There were three stenoses of > 50%, two were identified correctly, and the other showed as an area of signal void. Artifacts There was an area of signal loss over one disease-free common femoral vessel although this segment was well within the scanning plane. There were two areas of signal voids over the common iliac arteries. One was probably the result of adjacent bowel movement, as the vessel could be seen satisfactorily on some of the unprocessed images. An area of signal void was seen at the site of a Palmaz stent in the common iliac artery on the MIP images; on review of the individual sections, although the patency of the artery could be established, there was significant reduction in the apparent diameter of the vessel throughout the length of the stent when compared 1107
UM Sivananthan, J P Ridgway, K Bann et al
(b)
Figure 5. MR angiogram showing aortic occlusion in the maximal intensity projection display (a), and the unprocessed individual section revealing a thrombosed aortic aneurysm as the cause of the occlusion (b).
with the conventional films. In three patients the iliac, and in two the common femoral arteries, were outside the scanning planes. Artifacts owing to pulsatility (MR ghosting during cardiac pulsations) were seen as stepped variation in intensity in eight external iliac, and in all but two common femoral segments. On these instances, the vessel diameters were 5 mm or less on conventional angiograms. Discussion Early techniques for magnetic resonance angiography (MRA) of the lower extremities used cardiac gating and relatively long acquisition times (Meuli et al, 1986). The recent MRA techniques employ thin slices acquired in two dimensions (Gullberg et al, 1987; Edelman et al, 1989; Keller et al, 1989) or in three dimensions (Laub & Kaiser, 1988; Lenz et al, 1988). These methods produce vessel contrast without cardiac gating either by phase encoding as in rephased-dephased subtraction method or by flow related enhancement. FLASH (Fast Low Angle Shot) is an imaging sequence based on the gradients echo imaging method. Turbo-FLASH is merely an extension of this technique. Recent improve1108
ments in hardware have made very rapid ramping of the gradients possible enabling the TR (4-8 ms) and TE (2^4 ms) to be very short. 7j-weighting and tissue contrast can be varied by altering the flip angle. A larger flip angle achieves more Tx -weighting but in order to improve signal to noise ratio it necessitates a longer TR which in turn increases the acquisition time. However, this allows much thinner slices to be obtained in a multislice sequence. Thinner slices are necessary for the visualization of the iliac vessels, we selected eleven 5 mm slices to study first pass enhancement which requires an acquisition time of 19 s, a reasonable length of time for a single breath hold. We displayed the resulting images after postprocessing using the MIP algorithm, using the best slices from each acquisition. The MIP can be displayed in the same orientation as the conventional angiogram and is suitable for displaying tortuous vessels i.e. the iliac arteries. Because the MIP algorithm picks out only the highest intensity voxel encountered in the projected ray, it gives excellent contrast between the enhanced flowing blood and the stationary surrounding soft tissue. The contrast is maintained even when many sections are The British Journal of Radiology, December 1993
Fast MR angiography using turbo-FLASH sequences
Figure 6. MIP display showing the true extent of the right common iliac aneurysm (arrow) and infra-renal aortic stenosis (arrow head).
included in the MIP as long as the intensity of the blood is 2 standard deviations above the average background noise. Repeated performance of this process with slight alteration in the projection angle and then displaying them in a cine loop allows vessels to be seen without super-imposition. This allows more accurate evaluation of the aortic bifurcation, as seen in three of our patients who had significant posterior plaques which were not detected by conventional angiography. Vol. 66, No. 792
Artifacts can occur using MIP when thick slices are imaged, especially when the vessel intensity is approximately 0.5 standard deviations or less above the background intensity (Anderson et al, 1990). This can result in the vessel vanishing completely after postprocessing as fluctuations in background intensity can exceed the value of the vessel. When a vessel seems to be absent the cause could be occlusion, slow inflow or turbulence. Referring to individual sections may be useful in determining the correct cause. The MIP image can be improved by projecting together only a small selected number of sections at a time or by clipping the 3-dimensional data set to include just the area containing vessels of interest before projection thus minimizing the background. The reduction in the vessel width on MR images is owing to a combination of slower laminar flow along the vessel wall and also to partial volume effect of the marginal pixels within the lower intensity background. In the larger vessels the widest diameter variation (50%), when compared with conventional angiography, occurred in an external iliac segment owing to increased background signal intensity, probably caused by adjacent bowel movement. This discrepancy was reduced to 10% when only selected images were processed. Sites of all arterial occlusions were correctly identified by MR angiography. In the three aortic occlusions, not only their lengths were accurately determined, but also in two patients aortic aneurysms were identified as the cause of infra-renal aortic occlusions and in another patient a thrombosed normal diameter vessel could be clearly seen on MR angiography. Compared with the aortograms the MR images in addition to showing the patent lumen, demonstrated the nature and the true extent of the aneurysms. If lesions of 50% or more stenosis were taken, there was 87.5% and 70% specificity for aortic and iliac lesions, respectively, with a sensitivity of 100% for both. Three severe stenoses (two iliac, one femoral) were seen as areas of signal void. Typically arterial stenoses are accompanied by local turbulence which leads to decreased signal, this relative signal loss is amplified by MIP and can result in possible overestimation of vascular stenoses or non-visualization of the patent poststenotic vessel. Slow laminar flow along the vessel wall and the partial volume effect could be contributing factors. Most common artifacts encountered resulted from pulsatility, and were seen as stepped variation in intensity. This was apparent in eight external iliac and in all but two common femoral segments. The arterial blood flow in the aorta and the iliac arteries are so pulsatile that the flow reverses, causing the velocities to be small for a significant fraction of the cardiac cycle, resulting in a variation in vessel signal intensity between adjacent slices (de Graff & Groen, 1992). The invisibility of the very small vessels may be influenced by the slice thickness and by partial volume effect. Bowel movement still remains a cause for apparent signal loss, as seen in one 1109
UM Sivananthan, J P Ridgway, K Bonn et al
of our patients, selective processing of the slices significantly reduces this artifact. In a patient who had a patent Palmaz iliac stent there was signal loss on the MIP images and the unprocessed images throughout the length of the stent. This is owing to local artifacts induced by the metallic stent. When an apparent contour deformity of the vessel is noted, a distinction must be made between, mural plaque, turbulence, which may be present in the absence of underlying disease, and flow separation. Occasionally these are better differentiated by referring to the individual sections. Atheromatous plaques appear dark on gradient echo sequences (Kim et al, 1990) and can be easily distinguished from the bright flowing blood. Renal outlines were remarkably well seen in all 17 patients, this is because the highest signal is represented byflowingblood and structures which display a marked reduction in Tx after Gd-DTPA, such as the kidneys. The region of the body extending from the aortic bifurcation to the iliac bifurcations is important in assessing peripheral vascular disease of the lower extremities. This region is not suited for the rephased-dephased subtraction MRA method because of its large volume, the prominence of fat in many patients, and the visceral motion from respiration and peristalsis. The MRA technique of inflow signal enhancement using sequential 2-dimensional slices can give better results in the pelvis (Wendt et al, 1990), however, the vessel signal to noise ratio is poor and a major disadvantage of the above method is that, the number of slices needed to cover the abdominal aorta and the iliac arteries is rather high, thereby necessitating multiple acquisitions and long scan times. The turbo-FLASH technique we employed somewhat overcomes the above disadvantages. It is fast, it inherently has a wide coverage of the aorta and the iliacs as the scans are obtained coronally, and, as the scans are obtained after Gd-DTPA enhancement venous overlap
1110
is of no consequence. Although Gd-DTPA was given as a rapid bolus over a period of 2-3 s, it was well tolerated with no adverse reactions. This technique promises to be an alternative non-invasive method for assessing atherosclerotic aortoiliac disease. References ANDERSON, C M, SALONER, D, TSURUDA, J S ET AL, 1990. Artifacts in maximum-intensity-projection display of MR angiograms. AJR, 154, 623-629. DE GRAFF, R G & GROEN, J P, 1992. MR angiography with pulsatile flow. Mag. Res. Imag., 10, 25-34. EDELMAN, R R, WENTZ, K U, MATTLE, H ET AL, 1989. Projection arteriography and venography; initial results with MR. Radiology, 172, 351-357. GULLBERG, G T, WEHRLI, F W, SHIMAKAWA, A & SIMONS, M A, 1987. MR vascular imaging with a fast gradient refocusing pulse sequence and reformatted images for transaxial sections. Radiology, 165, 241-246. HAASE, A, 1990. Snapshot FLASH MRI. Applications to Tx, T2 and chemical-shift imaging. Mag. Res. Med., 13, 77-89. KELLER, P J, DRAYER, B P, FRAM, E K, ET AL, 1989. MR angiography with two-dimensional acquisition and threedimensional display. Radiology, 173, 527-532. KIM, D, EDELMAN, R R, KENT, K C ET AL, 1990. Abdominal aorta and renal artery stenosis: evaluation with MR angiography. Radiology 174, 727-731. LAUB, G & KAISER, W, 1988. MR angiography with gradient motion refocusing. JCAT, 12, 377-382. LAUB, G, 1990. Displays for MR angiography. Mag. Res. Med., 14, 222-229. LENZ, G W, HAACKE, E M, MASARYK, T J & LAUB, C, 1988. Inplane vascular imaging: pulse sequence design and strategy. Radiology, 166, 875-882. MEULI, R A, WEDEEN, V J, GELLAER, S C ET AL, 1986. MR gated subtraction angiography; evaluation of lower extremities. Radiology, 159, 411-418. WENDT, R E, NITZ, W, MORRISETT, J D & HEDRICK, T D, 1990. A technique for flow-enhanced magnetic resonance angiography of the lower extremities. Mag. Res. Imag., 8, 723-728.
The British Journal of Radiology, December 1993
