Magnetic resonance angiography and computed tomography angiography for peripheral arterial disease (2024)

Abstract

Introduction

The most common indication for the investigation of lower limb arterial disease is for those patients affected by atherosclerotic occlusive disease presenting with intermittent claudication or critical lower limb ischaemia, i.e. rest pain and tissue necrosis. Although these represent part of the spectrum of peripheral arterial occlusive disease, these scenarios are usually surprisingly distinct and patients presenting with intermittent claudication do not usually progress to critical ischaemia. Other disease processes affecting the lower limb arteries that may require radiological investigation include aneurysmal disease (especially of the popliteal arteries), thromboangiitis obliterans (also known as Buerger's disease, an inflammatory disease of the small and medium sized vessels, more common in men and related to smoking), cystic adventitial disease (a rare disorder of mucin-secreting cells in the arterial adventitia leading to cyst formation in the arterial wall that results in luminal compromise), the popliteal entrapment syndromes and trauma.

Historically, detailed examination of the peripheral arteries necessitated invasive arteriography and opacification of the vessels by iodinated contrast agents, with the inherent risks of arterial puncture, potential allergic and nephrotoxic effects of contrast media, and the hazards of ionising radiation itself for patients and staff. With the incidence of pseudoaneuysms and embolic events alone quoted as high as 1% and 0.5%, respectively, it is little wonder that the focus of attention in vascular imaging has turned to non-invasive techniques, of which magnetic resonance angiography (MRA) and computed tomography angiography (CTA) are now at the forefront.

Computed tomography angiography (CTA)

Background and technique

Since the introduction of helical CT into clinical practice in 1991, and the demonstrated ability to capture images of an arterial territory during first-pass contrast medium passage [1], CT has gained wide acceptance as a non-invasive technique for imaging vascular disorders. Whereas scanning with single-slice systems was limited to the assessment of focal abnormalities in a single vascular territory, such as aneurysms and abdominal bypass grafts, the advent of multidetector-row CT angiography (MDCTA) has extended the length that can be assessed. Indeed, still using a single bolus of contrast medium and without compromising spatial resolution, the faster acquisition times afforded by MDCT mean that it is now practicable to image the abdominal aorta and the lower limb arteries down to the feet. MDCTA has been found to be a useful tool in operative planning, as the display both of the vessel lumen and the mural calcifications helps to determine which vessels are suitable for cross-clamping and graft anastomoses.

Peripheral CTA requires no special patient preparation other than ensuring that the patient is well hydrated as per local institutional guidelines to minimise the risk of contrast medium nephropathy, as this patient population often has coincident renovascular disease and many are diabetic. Indeed, often the most difficult part of patient preparation for CTA is ensuring that patients are not administered any oral contrast agent prior to the procedure! This is important, as bowel opacification significantly degrades the quality of post-processed images.

Technical factors

CTA relies on thin-section helical imaging and requires optimum timing of contrast medium to ensure adequate opacification of the whole vascular territory. Although the exact technique used depends on the type and capabilities of the scanner, the aim is to optimise spatial resolution along the z-axis by choosing the smallest slice thickness possible that will still enable the anticipated scan range to be covered within a reasonable time during which the vessels are opacified (20–40 s). For example, with a 4-slice scanner capable of a 0.5-s gantry rotation time, a collimation of 2.5 mm (giving an effective slice width of 3 mm) with an effective pitch of 1.5 (i.e. table feed of 15 mm per rotation) will cover a range of 60 cm in 20 s and 120 cm in 40 s. Images are routinely reconstructed in an overlapping fashion (usually 50% overlap) to improve the quality of reformatted maximum intensity projection (MIP) images. A slice thickness of 1 mm will afford greater spatial resolution in the z-axis, however when used with an effective pitch of 1.5 (i.e. table feed of 6 mm per rotation) on a 4-slice scanner this will cover only 30 cm in 25 s and 60 cm in 50 s, with the feasibility of longer coverage being unlikely owing to tube heating effects. With MDCT scanners using more than four rows, the z-axis spatial resolution can be improved; for example, using a collimation of 1 mm for peripheral CTA with a 16-detector row scanner, a full peripheral run-off examination of 120 cm can be achieved in 25 s.

Contrast medium technique

Contrast medium concentrations of 320–350 mg I ml^–1 are usually used; although there are advocates of the newer 400 mg I ml^–1 contrast media, there is little published evidence of their benefit. Whilst higher concentrations can be used to reduce flow rates, the degree of enhancement is actually dependent upon the amount of iodine arriving in a region per second, not the concentration of iodine contained per ml of contrast medium. Contrast medium injection can be either monophasic or biphasic. A monophasic bolus will lead to a gradual increase in intravascular contrast medium concentration, with a relatively fast “drop-off” after peak enhancement. A biphasic injection will achieve a more constant contrast medium concentration, as a rapid initial delivery of contrast medium is followed by a second slower bolus to maintain enhancement, and this technique produces more uniform aortoiliac enhancement, potentially aiding the evaluation of small peripheral arteries [2]. The timing of scan commencement after peripheral intravenous injection is determined either by a test bolus injection of a small volume of contrast medium or, alternatively, a “bolus tracking” technique can be employed whereby angiographic acquisition is triggered when a threshold Hounsfield unit (HU) attenuation value is exceeded in the aorta (see Figure 1). The bolus tracking method was recently compared with test bolus injection in cardiac CTA by Cademartiri et al [3] and found to yield more hom*ogeneous enhancement; it is as yet unknown whether this improves peripheral MDCTA quality, but it does reduce examination times.

Whether a monophasic or biphasic technique is used, the contrast medium should ideally be followed by a bolus of saline (approximately 50 ml), which affords longer enhancement; this may also allow a reduction in the amount of contrast medium required, usually by approximately 30%, as it flushes out otherwise “wasted” contrast medium from the upper limb venous system [4]; however, the dual-head contrast pump injectors required for this are still relatively uncommon in most UK departments.

CTA visualisation

The images produced by CTA must be viewed and manipulated on a workstation, with inspection both of the base axial two-dimensional (2D) images and the reconstructed 3D images to prevent misdiagnosis. Review of the full set of original axial sections is particularly important for CTA [5, 6], and an important point to note for CTA evaluation is that window settings must be altered when evaluating calcified vessels, otherwise plaques tend to appear larger than they actually are. Widening the windows from the usual CTA level of around 150 WC ± 250 WW to 200 WC ± 1000 WW is recommended (see Figure 2).

Simple multiplanar reformation (MPR) as used for image review in CT of other body areas can be employed for overview of vascular anatomy and pathology, but standard MPR in a straight plane clearly gives a limited overview of curved and branching structures such as vessels. Curved planar reformations (CPRs) along the vessel long axis can be more useful but may be equally problematic, as “pseudo-stenoses” can be created if there is any deviation from a vessel's midline during editing (see Figure 3).

MIP is the simplest vascular post-processing method; this displays the maximum voxel value encountered along the projection path through a stack of 2D images as the MIP image pixel. Overlap of vessels can obscure detail, necessitating the use of targeted subvolume MIPs and interactive MPR of the 3D data set to allow accurate evaluation of stenoses and vascular relationships. Subvolume MIP (e.g. the use of a limited 20 mm slab) is a valuable tool, however care must be taken during editing to preserve image quality. The problems with MIP imaging for CTA are that the superimposition of calcifications on the vessel lumen and the presence of obscuring bone make standard MIP without any editing of limited use. Although MIP after removal of the bony skeleton may be useful for CTA in providing an anatomical overview, e.g. pre-operatively, accurate evaluation of stenoses on CTA MIPs may be impossible in the presence of calcification (Figure 4). The problem of obscuring skeletal structures and calcifications as well as stents can potentially be overcome by the use of volumetric subtraction techniques whereby a mask image 3D volume prior to intravenous contrast medium administration is subtracted from the CTA data set [7]. CTA subtraction is a new technique with software as yet not widely available, and this CTA subtraction method does have the obvious drawback that a whole extra CT data set needs to be acquired, doubling the radiation dose, whilst any movement between acquisitions may impair image quality though misregistration of calcified plaque, stents etc.

Other 3D rendering techniques are useful for overview anatomy; surface shaded display (SSD), where a particular HU value is assigned as an isosurface, is particularly prone to exaggeration or obscuration of pathology such as stenoses and has largely been superseded by volume rendering techniques (VRTs). VRT assigns different colours to tissues with defined HU value ranges and the opacity of these colours can be varied. This technique is particularly valuable for CTA, as useful information can be gained without tedious editing of the CTA data prior to rendering (e.g. bone removal; Figure 5); additionally, CT HUs are better standardised than MRI SI values so that pre-set VRT settings can be utilised, whereas with MRI VRT tends to require more individual “tweaking” to obtain good images. The problems with VRTs are that manipulation of images can be difficult and time consuming, and editing such as bone removal is still required for the best image quality (Figure 6). The technique is also demanding of workstation hardware and with the rapid increase in CTA data set size, it seems that this problem is only likely to get worse [8]. Additionally, subtle changes to the VRT parameters can yield marked changes to the rendered images, and the technique tends to mask branch vessel detail compared with MIP images (Figure 7) that better preserve fine details and stenoses compared with VRT or SSD [9, 10]. VRTs also still struggle with calcifications; particularly problematic are those lesions (arterioliths) often found in the external iliac/femoral segment that impinge into the lumen and whilst calcified, are not as dense as mural plaque, thus they are actually of similar attenuation to the contrast-enhanced lumen, are rendered effectively invisible against the lumen and can be overlooked [6].

Therefore, when viewing overview MIP or VRT images with CTA contiguous data, the whole anatomy from abdominal aorta to the feet can be seen. However, for critical diagnostic review, CTA renderings should be divided into sections (e.g. aortoiliac, iliofemoral, femoropopliteal and popliteal/crural), this limitation being a function of the display size of workstation monitors. For example, the 120 cm or so scan range from aorta to feet when split into three or four overlapping regions of 35–50 cm will then correspond to a pixel size of 0.7–1 mm and the display of each section will be at “true” resolution, whereas if the whole length is displayed at once the detail will be lost as pixels are averaged.

Clinical applications of CTA

Abdominal aorta

CTA has been advocated as an alternative to conventional angiography in the pre-operative evaluation of abdominal aortic aneurysms, particularly prior to endovascular aneurysm repair (EVAR). Not only is CTA less invasive, it is more accurate in predicting aneurysm size and can demonstrate mural thrombus and ulcerations within aneurysms, inflammatory changes, perianeurysmal blood secondary to contained rupture, and the presence of any co-existing non-vascular intra-abdominal disease. The use of helical CTA allows the application of 3D multiplanar rendering techniques, resulting in greater accuracy in aneurysm and lumen measurement for EVAR planning. However, the role of CTA in the accurate demonstration of juxta/suprarenal aneurysmal extension remains unclear. In addition, assessment of the presence, position and patency of aortic side-branches by CTA is felt by some to be inadequate.

Peripheral arteries

The preliminary study by Rieker et al [11] evaluated 50 patients prospectively against intra-arterial digital subtraction angiography (IADSA) using a single-slice scanner, covering only the region from the groins to the calves (excluding pelvic and pedal arteries), and found reasonably good accuracy for femoral, popliteal and tibioperoneal trunk occlusions. Despite the advance that MDCTA appears to represent over single-slice scanning and its apparent increasingly established role in angiographic imaging of the lower limbs, there are actually few clinical reports in the literature compared with MRA and no meta-analyses to help define the place of this technique compared with contrast-enhanced MRA. Rubin et al [12] described their initial experience using a 4-slice scanner assessing arterial inflow and run-off opacification in 24 patients with mixed disease (symptomatic lower limb occlusive arterial in 19 patients and aneurysmal disease in 5 patients). They used a test injection with an aortic region-of-interest placed to produce a time–attenuation curve and hence to determine scan delay for a contrast bolus of 180 ml at 3.6 ml s^–1, with CTA performed using a 2.5 mm slice thickness. The axial sections obtained were reconstructed at 1.6 mm intervals using a modified 180° linear interpolation algorithm and the angiograms were processed into MIPs, CPRs and volume renderings. In 18 of these patients, conventional angiography was required on clinical grounds within 3 months of the CTA, following which two radiologists independently reviewed the studies. Their results showed CTA to be a robust and reliable alternative to conventional angiography, with 100% concordance between the two techniques for arterial opacification, CTA depicted 26 additional segments not adequately demonstrated by conventional means owing to improved arterial opacification distal to occlusions. This study also showed that the radiation dose was 3.9 times greater with conventional angiography than with CTA. Other papers by Beregi et al [13] and Soto et al [14] have demonstrated additional advantages of CTA, such as increased detection of popliteal artery aneurysms and higher pick-up rates for arterial injuries in trauma patients.

Catalano et al [15] published one of the largest CTA studies to date, considering 50 patients examined with 4-detector row CTA. Again this was a mixed population and 27 of these patients had grade I or II disease whilst the other 23 had grade III lower limb ischaemia. This group showed good agreement with consensus evaluation for imaged segments derived from three observers compared with DSA. However, CTA in this study did tend to overestimate stenosis degree. Unfortunately, other studies have evaluated even fewer patients; for example, while Ota et al [5] quote high sensitivities and specificities of 99–100% compared with DSA, they evaluated only 24 patients and these figures are for “detection of more than mild stenosis”. A recent study of 47 patients by Romano et al [16] reported an overall diagnostic accuracy of 94%; this study had good interobserver agreement, with kappa values of 0.8, but showed a tendency for poorer performance for calf vessels. The study by Heuschmid et al [17] evaluated a smaller number of patients (23), with accuracy above the popliteal trifurcation of 87.3% and distal to the trifurcation of 80.2%. Meanwhile, Mesurolle et al [18] assessed CTA in just 16 patients against transcatheter angiography, with overall sensitivity of 91% and specificity of 93%; however, segmental analysis in this study found a sensitivity of only 43% in infrapopliteal arteries and a specificity of 86%. The most recent prospective study by Edwards et al [6] using a 4-slice CTA technique in 44 patients found disappointingly poor sensitivities of 79% and 72% for two observers assessing for treatable lesions >50% stenosis. This poor sensitivity was due to poor delineation of short segment stenoses through partial volume averaging as well as poor contrast discrimination between calcified arterioliths and the enhanced lumen. Whilst MDCTA did show an increase in patent segments beyond long occlusions, this group concluded that “MDCTA using 4-slice machines is insensitive to detecting significant arterial stenoses in the lower limb arteries”. Clearly further work with thinner collimation using MDCT scanners with more detector rows (16-slice and greater) is required.

Pitfalls and problems with CTA

Although the studies are small and the initial clinical results encouraging, there are several problems and challenges still facing peripheral vascular CTA. First, CTA by necessity requires the use of X-ray ionising radiation; whilst it is less than IA-DSA and modern scanners have dose-reducing technology, it is still a significant radiation dose. It can be argued that in the more elderly population predominantly afflicted by peripheral vascular occlusive disease the radiation burden is of little consequence, but this is not the case in younger patients who comprise a growing proportion of this population, with accelerated vascular disease requiring investigation seen particularly in diabetics and those with renal disease. A perhaps more important problem is that of contrast medium nephrotoxicity [19–21], with contrast-induced nephropathy (CIN; defined as a creatinine rise of >0.5 mg dl^–1 (44.2 μmol l^–1) or >25% above baseline) occurring in 2% of general patients but up to 50% of high-risk patients such as those with diabetes and impaired renal function. The importance of CIN is highlighted by the fact that it is associated with a mortality of up to 34% in those inpatients who develop this complication.

Although multiphasic injection strategies have been shown to produce greater uniformity of enhancement throughout the scanning period, they have not been validated for scan durations of over 30 s [2]. Additionally, substantial discrepancies in circulation times between the legs can occur in the setting of lower limb occlusive disease and contrast medium timing should take into account any pre-existing knowledge about the type of disease being imaged. Whilst a unilateral superficial femoral artery occlusion with good collateralisation in an otherwise well claudicant patient will cause only a minor delay in contrast enhancement more distally, the presence of ectatic or aneurysmal vessels may delay the distal flow of contrast medium to such a degree that vessels “downstream” are insufficiently opacified as the contrast bolus is “outrun”; this may become even more problematic as the next generation of faster MDCT scanners are employed. The patient's cardiovascular status also influences the rate of contrast medium travel and should be taken into account when setting up scan parameters. When severe asymmetries exist, the diagnostic quality of an initial conventional angiogram can be limited, necessitating repeated contrast medium injections. Owing to the volume of contrast medium injected at CTA and the risks of CIN, multiple injections are precluded, making it critical that both legs are sufficiently opacified. Rubin et al [12] found maximum right–left differences in HU values of 75 above the calf and 96 below the calf but did not feel that image interpretation was adversely affected in their cohort, predominantly claudicants.

Overlying venous opacification has long been recognised as a limitation in the display of arterial structures on 3D views, particularly when examining the carotid and renal arteries. Although lower limb veins tend to be best visualised 3.5 min after contrast medium injection, some superficial veins are seen substantially earlier, particularly in critical ischaemia where there is often rapid arteriovenous shunting. Whilst this is not a major problem when viewing stacked axial sections, the quality of 3D arterial reformations is severely compromised and CTA studies to date have not focused on patients with Fontaine class III & IV critical ischaemia in whom this will be most problematic.

Lastly, whilst the visualisation of calcification may be potentially useful, it can also be highly problematic in terms of lumen visualisation where the superimposition of calcifications on the vessel lumen can make interpretation near impossible and as described above, post-processing CTA data sets tend to be more complicated and the results more operator-dependent than for MRA.

Magnetic resonance angiography (MRA)

Background

All MRA techniques have one principle in common — maximisation of signal differences between stationary tissue and blood in the vessel lumen. Developed in the 1980s, MRA was first used to visualise the carotid and intracranial arteries using the flow-sensitive imaging techniques of time-of-flight (TOF) and phase contrast (PC) imaging.

TOF-MRA

TOF-MRA produces flow-related enhancement by manipulating the longitudinal magnetisation of static tissue, preventing their T1 recovery by using rapid repetitive radiofrequency (RF) pulses to produce steady-state saturation and thus low signal. Meanwhile, contrast is provided by the in-flowing unsaturated blood producing high signal within the vessel lumen. Images can be acquired as 2D, 3D or as a hybrid technique known as 3D Multislab, with the 2D method being the most established for peripheral vascular MRA. Contiguous slices are acquired sequentially and stacked to cover the desired imaging volume. For peripheral run-off imaging, saturation bands are applied caudad to the imaged slice to suppress unwanted signal from caudocranial venous flow. These saturation bands can themselves cause signal loss in very pulsatile flow in the iliac and femoral arteries, as arterial blood may move completely through the imaging slice, enter the caudal saturation band and then return to the imaging slice within the TR imaging time and thus reduce signal. This can be ameliorated either by distancing the saturation bands or by using cardiac/pulse gating (which even further prolongs the imaging time), but this is now really only of historical interest as several constraints limited the use of 2D TOF-MRA as a routine clinical tool. 2D TOF-MRA performed well in trials [22], in particular being more sensitive for depiction of run-off vessels than conventional arteriography; however, while it is very sensitive to flow that enters the field-of-view (FOV) perpendicularly, in-plane flow that traverses the imaged volume can be saturated along with stationary tissues resulting in signal loss. Furthermore, imaging vascular structures deep in the body such as the renal vessels was hampered by poor signal return and motion artefacts and moreover, areas of disturbed or slow flow, such as at the site of arterial stenoses or within aneurysms, can cause complex flow-related artefacts [23]. Ultimately, the length of time for image acquisition of these large regions was prohibitive, with comprehensive studies taking an hour or more, although limited evaluation of the distal tibial arteries and feet remained an indication where conventional angiography had difficulty until the advent of contrast-enhanced MRA.

PC-MRA

PC-MRA alternatively utilises the changes in phase (direction) of moving protons relative to stationary tissues produced either by physiological conditions such as velocity changes between systole and diastole, or by using a bipolar gradient to encode flow velocity. 2D or 3D images are produced where static tissue displays very low signal and moving tissue has a higher signal proportional to its speed of flow. PC-MRA therefore quantifies both magnitude and direction of flow, and flow rates can be calculated from voxel velocity data. Whilst PC-MRA has been applied to peripheral run-off evaluation [24], it has never been found useful clinically owing to the long acquisition times as per TOF-MRA; however, it has a potential role for problem-solving and stenosis quantification in iliac vessels and across stents.

Contrast-enhanced MRA

The status of body MRA has been radically altered with the advent of gadolinium 3D contrast-enhanced MRA (3D CE-MRA) and it is now at the forefront of non-invasive vascular imaging. Indeed CE-MRA has all but replaced 2D TOF-MRA owing to the speed of the technique and its insensitivity to spin dephasing, with resultant increased accuracy and confidence in assessment of stenoses, thus 2D TOF-MRA has been relegated to providing basic localiser images for planning CE-MRA studies. Gadolinium-based paramagnetic contrast agents increase the blood signal by causing a profound shortening of its T1 relaxation time. A fast T₁ weighted gradient echo pulse sequence is used, specifically designed to produce an image with low background signal through RF spoiling, the result being akin to a conventional angiographic image (although in 3D) as there is little or no flow-related information. As vascular contrast depends only on the T1 shortening effect of gadolinium, CE-MRA is not susceptible to the in-plane flow saturation effects seen in TOF-MRA. Imaging can therefore be performed in the plane of the vessel of interest, not merely perpendicular to it, which in turn greatly increases the practicable FOV, reducing acquisition time. The first publications regarding coronal acquisition contrast medium infusion MRA were specifically for the investigation of advanced aortoiliac disease as an alternative to conventional angiography [25], this paper described the use of a coronal 2D gradient-echo technique (turbo-FLASH) with infusion of gadolinium contrast, whilst Prince et al described a 3D gradient-echo technique with coronal acquisition during gadolinium contrast infusion. These early 3D-MRA studies employed pulse sequences that took several minutes to acquire, and both hardware (gradient performance) and software improvements were required to allow faster acquisitions.

To produce good vascular contrast in CE-MRA, a high concentration of gadolinium contrast agent must be achieved in the vessels of interest to reduce the T1 of blood to less than that of adjacent fat (i.e. less than 270 ms for a 1.5 T system), which in practice means an intravascular gadolinium concentration of greater than 1 mM. The higher the injection rate, the higher the achieved concentration for any given contrast medium dose. However, this will mean a shortened bolus profile and thus also necessitates faster data acquisition sequences. Large areas can be imaged using non-breath-hold techniques, although motion artefact from respiration reduces diagnostic accuracy for renal/visceral branch vessels in the abdomen. Breath-hold techniques are more desirable for the abdominal aorta and can be used for a volume acquired in 30 s or less. Whilst the short TR (3–7 ms) and TE (<2.5 ms) required to image a clinically useful volume in a breath-hold necessitates fast gradient subsystems, these are now commonplace on modern MRI installations in the UK. The use of a stepping-table technique, analogous to that used in conventional angiography, has since overcome the initial problem of inadequate FOV by acquiring sequential FOVs (stations) in a bolus chase fashion, allowing visualisation of aortoiliac, femoropopliteal and tibial segments [26].

Peripheral CE-MRA technique

No special patient preparation is required for CE-MRA, although there could be a case made for avoidance of blueberries, pineapple juice, chocolate and other foods rich in manganese as they can produce high signal on T₁ in the bowel [27]. Prior to the actual scan, the patient should be instructed on the importance of breath-holding for imaging that includes the abdominal vessels, and an intravenous cannula placed in a suitable arm vein. As for any MRI examination, the initial imaging is to acquire localising scans of the region-of-interest to best plan the MRA itself. For a CE-MRA study it is usual to acquire a rapid, low resolution TOF-MRA to allow exact planning of the coronal CE-MRA stations. In peripheral MRA it is important to acquire mask image volumes prior to contrast enhancement, which serve as a reference enabling subsequent image subtraction to increase the contrast-to-noise ratio (CNR); subtraction of regions of high signal pre contrast (such as subcutaneous and marrow fat) from the post-contrast images improves vascular image quality (Figure 8) [28]. These mask images are also available for review to allow detection of local areas of high signal pre contrast medium injection (e.g. thrombus containing methaemoglobin) or local susceptibility artefact from metal (surgical clips etc.), which may otherwise confound subsequent image interpretation. Whilst image subtraction improves aortoiliac station images, it is particularly crucial for the tibial station, as the image scaling effect of high signal from contrast medium is negligible with the small crural vessels and both the subcutaneous and marrow fat are in closer proximity to the vessels. Image subtraction is usually performed “on-the-fly” during image reconstruction and complex number subtraction is usually implemented as it provides higher CNR.

Contrast technique for CE-MRA

Optimum timing of the gadolinium contrast bolus is necessary to ensure that the peak concentration in the vessel of interest, and therefore the most profound T1 shortening of blood, coincides with the collection of the central lines of k-space, as these low spatial frequencies determine contrast. In the original studies with conventional k-space encoding and relatively long acquisitions, this was achieved empirically using a “best guess” technique; subsequently bolus timing was shown to be more reliable and the central part of the sequence was timed to coincide with peak enhancement. Now most systems are supplied with automated triggering techniques and/or fluoroscopic monitoring (Figure 9) and employ centric type CE-MRA pulse sequences whereby the central lines of k-space are acquired at the start of the acquisition, allowing a simpler determination of the start of the pulse sequence, as the scan delay simply equals the time to enhancement. The only downside to the use of centric sequences is that if an acquisition is started too early then a “ring” truncation artefact may result, with dark central stripes in the vascular lumen (Figure 10); this problem is ameliorated by the use of specially tailored centric acquisition schemes designed for CE-MRA such as contrast-enhanced timing-robust angiography (CENTRA) [29].

Bolus timing involves the use of a test injection of gadolinium contrast, with the area of interest (e.g. aorta) repetitively imaged and the time of peak vascular enhancement ascertained to the nearest second from injection similar to CTA test bolus techniques. Only a small test volume of gadolinium contrast is required (e.g. 2 ml), however it must be followed immediately with a saline bolus (30 ml) to flush it from the arm veins and to best mimic the subsequent full contrast bolus dose. A period of prior hyperventilation should be combined with the scan delay to ensure successful breath-holding.

As with any test dose, an equivalent volume of saline should immediately follow the main bolus dose at the same injection rate to ensure that all the contrast medium leaves the upper limb veins. The whole injection process is simplified by the use of an MRI-compatible pump injector, which both ensures reproducibility and allows the medical and radiographic staff to remain in the control room. On average, a total dose of 20–30 ml standard gadolinium contrast (approximately 0.2 mmol kg^–1) is used for a single-station study. This is increased to approximately 45 ml to perform multistation imaging of the aorta and peripheral arterial run-off, with a biphasic injection protocol using a contrast medium rate of 1.5 ml s^–1 for the first phase then dropping to 0.8 ml s^–1 followed by saline flush. A smaller or larger volume may be required depending on patient size and estimated cardiac output.

Optimising peripheral CE-MRA

The main problems with peripheral run-off CE-MRA involve the classic MRI trade-offs of spatial resolution vs signal-to-noise ratio (SNR) vs image acquisition time. Standard stepping-table CE-MRA is often sufficient for imaging patients with simple claudication. However, for critical ischaemia high spatial resolution is particularly important for the small tibial arteries, yet this is also the region where venous contamination is most likely in patients with Fontaine class III & IV disease (Figure 11), although it can even be seen in patients with simple claudication (Figure 12). The problem here is that arteriovenous transit time paradoxically shortens with the opening up of microvascular shunts and so the time window between arterial and venous enhancement is reduced. This has led to several different strategies to optimise tibial arterial imaging in these patients, each having their own advantages, as detailed below.

The simplest method, as described by Morasch et al [30], is to ensure best accuracy for timing of acquisition for the calf vessels by imaging them first with a dedicated injection of contrast medium and utilising a centric-encoded sequence designed for venous suppression, i.e. elliptic centric and variants such as CENTRA [29]. The aortoiliac and femoral levels are then imaged with standard stepping-table technique using a second subsequent contrast medium injection. This has the advantage in that an increased spatial resolution acquisition can be used at the crural level appropriate to the smaller calibre tibial arteries and a phased array coil employed to boost SNR (Figures 12–14).

Another strategy is to again image the calf vessels first, but instead of static 3D imaging, to use thick slab 2D imaging with high temporal resolution, so-called MR-DSA [31]. With MR-DSA, although the vessels are only seen in one projection plane and the in-plane spatial resolution may also be compromised, this is made up for by the gains in temporal resolution whereby the vessels are seen to fill and collateral pathways are appreciated, whilst venous contamination is only seen as part of the dynamic sequence as a late feature that does not obscure arterial information.

Alternatively, the imaging time window itself at the calf can be prolonged by the use of venous compression techniques, which entails the use of cuffs around the thighs or upper calves inflated to subdiastolic pressures to obstruct venous return [32–34]. This has the effect not only of prolonging arterial transit and venous transit times but additionally increases the venous volume of dilution, therefore any contrast agent that passes from arterial to venous compartments during acquisition will be more dilute and hence less conspicuous.

The use of parallel imaging and phased array coils at all stations with variable acquisition parameters allows more tailored and rapid bolus chase such that venous contamination is minimised, e.g. moderately high spatial resolution is acquired for the aortoiliac station then the femoral segment is very rapidly acquired with both increased slice thickness and lower in-plane resolution, then the highest spatial resolution is used for the tibial station with a heavily centric acquisition such as CENTRA [35, 36]. Dedicated peripheral vascular phased array coils have the additional advantage that the pedal arch is consistently visualised compared with standard stepping-table techniques.

Lastly, with the use of parallel imaging and phased array coils, the tibial station can again be acquired first but, with the extra efficiency gained, high spatial resolution 3D time-resolved imaging is implemented as an improvement on 2D MR-DSA; this is effectively a hybrid of the techniques described by Morasch et al [30] and Wang et al [31] (Figure 15).

MRA visualisation

As for CTA, the images produced by MRA must be viewed and manipulated on a workstation with inspection both of the base 2D images (usually coronal partitions from a 3D acquisition) and the reconstructed 3D images to prevent misdiagnosis. As well as review of the full set of original sections, it is important that mask images are reviewed if there is any concern about any unusual lesions that could relate to susceptibility artefact.

MIP is the most commonly used and simplest post-processing method; overlap of vessels such as the superior mesenteric artery can obscure detail in the abdomen, but this is less problematic for the lower limb arteries. To evaluate abdominal and tortuous pelvic vessels, use of thin section (e.g. 20 mm or 40 mm) subvolume MIPs may be helpful. In general, MIP imaging is a particularly valuable tool for MRA for which it usually suffices as the sole rendering technique required, since the CTA problems with superimposition of calcifications and bone do not arise.

As with CTA, MPR and CPR can also be employed for overview of anatomy and pathology, but these are seldom required; similarly, SSD and VRT are rarely needed, only being necessary where there is particularly tortuous anatomy that needs perspective shading to allow clarity of presentation.

The majority of peripheral run-off CE-MRA studies are acquired with automated “stepping-table” techniques; the different stations are acquired with a known degree of overlap and for presentation can be stitched together in software for review and hard-copy printing. However, for critical diagnostic review, evaluation of each station on the workstation is required, with similar concerns about display resolution as discussed previously for CTA.

Clinical applications of CE-MRA

Abdominal aorta

MRA can be used in the evaluation of aortic aneurysmal disease (Figure 16), stenoses, inflammatory changes [37] and dissection. MRA is particularly indicated in assessment of occlusive aortoiliac disease where the pulses are poor or absent and in those patients with grafts where invasive puncture is often problematic or even impossible (Figure 17). CE-MRA is also indicated in preference to CTA or conventional angiography in patients who either have renal dysfunction or are at-risk of renal impairment, since gadolinium contrast agents have little nephrotoxicity in the doses routinely used [38, 39]. The diagnostic performance of MRA in the aortoiliac segment is high for these large vessels, with many studies now having shown very high sensitivity and specificity compared with conventional invasive angiography (Figure 18).

Peripheral arteries

For the lower limb arteries the majority of the literature has focused on occlusive disease. MRA also performs well for popliteal aneurysms, arteriovenous malformations etc. (Figure 19), however it is not currently a first choice for assessment of arterial trauma. A particular advantage of MRA is the ability to discriminate conditions such as cystic medial degeneration by imaging the vessel wall, and popliteal entrapment syndromes by visualising compressing muscle slips and tendons [40]. CE-MRA has been shown to have better diagnostic power than duplex ultrasound for evaluation of occlusive disease and is equivalent to conventional angiography in terms of sensitivity and specificity (Figure 20); indeed, the CE-MRA literature is now relatively mature, with two meta-analyses published, both finding high accuracy for CE-MRA compared with DSA [41, 42]. Furthermore, in several studies specifically relating to chronic critical lower limb ischaemia, CE-MRA has been shown to outperform DSA in demonstrating an increased number of patent infrapopliteal vessels beyond occlusive disease [36, 43]. Given this accuracy and its lower cost compared with conventional arteriography, CE-MRA is clearly a highly cost-effective outpatient imaging technique that, when properly performed and interpreted, is sufficient on its own for planning peripheral bypass procedures.

Pitfalls and problems with MRA

As with any MRI technique, for certain patients this modality is not safe, specifically those who are pacemaker-dependent, have implanted cardiac defibrillators or other devices such as neurostimulators, and also for those with ferromagnetic intracranial aneurysm clips or retained ocular ferromagnetic foreign bodies.

Since CE-MRA requires specific prescribed imaging volumes to be acquired to cover the relevant anatomy, careful planning is required to ensure that the imaging volumes cover the vessels of interest fully. For example, the coronal imaging volumes may be limited in their anteroposterior coverage depending on number of partitions/slices, their thickness and system performance. For example, in the aortoiliac segment it is imperative that the volume coverage is checked to include both the femoral arteries fully anteriorly (as well as crossover grafts if present) as well as the most posterior parts of the iliac arteries in the pelvis, otherwise these segments will pass out of plane and not be evaluable. Assuming both correct contrast timing and volume coverage and notwithstanding the above comments on venous contamination, the main pitfall of MRA is in the use of appropriate spatial resolution for the expected size of vessel and pathology; if spatial resolution is not appropriate then stenoses may be underestimated. Research has shown that 3 pixels across the diameter of a vessel are required to accurately assess its diameter [44] and it thus follows that to accurately characterise a 50% diameter reduction stenosis then 6 pixels are required across the diameter of the parent vessel, i.e. to allow 3 pixels across the residual lumen to allow its accurate measurement. Hence, for an iliac artery of 6–8 mm diameter, a spatial resolution of approximately 1 mm is required, whilst for tibial arteries of 2–3 mm calibre then resolution should if possible be increased to the order of 0.5 mm.

The other major pitfall concerns artefacts from metallic structures; whilst vessels passing near hip prostheses and other surgical hardware may be affected by susceptibility, this is often surprisingly less of a problem than expected and is not as severe as on CT (Figure 21). If surgical clips lie directly adjacent to a vessel then artefact may compromise imaging, however the presence of the clip will often be inferred in that the vascular appearance produced will not look like physiological/pathological disease and artefact will be visible on the mask images (Figure 22) [45]. More important are the difficulties encountered imaging metallic vascular stents; the lumen of steel stents may be completely obscured though a combination of susceptibility artefact and RF shielding of the interior. Nitinol stents appear much less prone to these effects, as are those made from cobalt alloys and platinum, particularly so for those of larger calibre [46]. The RF shielding effects obscuring the interior of stents may be in part overcome by increasing RF flip angle.

In the future, it is likely that new techniques will afford higher image resolution, depicting even finer vascular detail. Specialised coils will allow both higher SNRs and faster image acquisition, and speed will help to eliminate venous contamination as a clinical problem. New contrast agents with higher specific relaxivity will help to preserve SNR with these small voxel sizes, and persistent vascular agents will also find a role. It may be then, after technological advances and appropriate clinical validation, that CE-MRA will completely replace conventional diagnostic angiography in an increasing number of clinical settings.

MRA vs CTA

Whilst the reliability and accuracy of CTA is clearly increasing, it is as yet difficult to determine exactly what role MDCTA will play given the advances being made in CE-MRA. There are reasons why CTA may become a compelling alternative to MRA in certain patient groups, for example some with co-existing coronary artery disease are increasingly having implantable pacemakers, resynchronisation devices and defibrillators precluding the use of MRI. Additionally, the simultaneous assessment of vascular calcifications achievable with CTA may have therapeutic relevance, particularly with regards to pre-operative planning of bypass procedures. However, as well as the obvious advantage MRA holds over CTA with respect to radiation dose, in addition the large number of patients with co-existing diabetes and renal disease can undergo investigation without risk owing to lack of nephrotoxicity from the contrast media. MRA also holds the advantage of simpler post processing, which can be largely automated, as well as the potential for multiphase examinations and time-resolved techniques.

Although much research is needed to address these questions, ideally in the form of large, blinded, prospective trials, we are only aware of one study to date that directly compares the two techniques. Willmann et al [47] compared CE-MRA and MDCTA of the aortoiliac and renal arteries in 46 consecutive patients, using IA-DSA as the standard of reference. In this study a total of 769 arterial segments were analysed for stenoses using a four-point grading system. The time required for performing 3D reconstructions and interpreting the images was also recorded for each modality. The sensitivity of MRA for detecting haemodynamically significant stenoses was 92.5% with a specificity of 99.5%. Sensitivity for CTA was 91.5% with a specificity of 99%. These differences between the two modalities were not statistically significant. The time required to reconstruct and interpret CTA images was significantly longer than that required for MRA, although more patients preferred CTA as their technique of choice.

Only with further research will it be seen whether CTA can fulfil the technological and clinical promises that came with its introduction in the late 1990s. However, as a technique that remains in its infancy, it already provides a useful addition to the ever-increasing range of modalities available for the investigation of peripheral vascular disease. Meanwhile, the advances in MRA gather apace, with higher spatial resolution, increasingly flexible examination protocols, new contrast agents, plaque imaging, time-resolved imaging and even the tantalising possibility of endovascular intervention becoming practicable in a radiation free environment.

© The British Institute of Radiology