Contrast-enhanced MRA at 3.0T

Although gadolinium based agents have an excellent safety record, reports linking gadolinium to nephrogenic systemic fibrosis have sparked renewed interest in "low-dose" CE-MRA and NCE-MRA [6–8]. In addition, low-doses of contrast also help reduce the costs of performing CE-MRA. Gadolinium chelates are paramagnetic compounds that shorten T1 and T2 relaxation times by disrupting spin-lattice and spin-spin interactions respectively. These effects of Gd on body tissues are relatively unaffected by increased magnetic field strength. Thus, although body tissue T1 relaxation times are increased at 3.0T, the T1 relaxation times of Gd-contrast agents remain relatively unchanged at higher magnetic field strengths. This results in noticeable increases in the blood pool-to-background contrast-to-noise ratio (CNR) compared to 1.5T. The increase in CNR at 3.0T can be used to improve the image quality using the same amount of contrast or to decrease the amount of IV contrast injected compared to a similar scan at 1.5T (Figure 2) [5, 9]. Tomasian et al. recently demonstrated that for 3.0T MRA of the supraaortic arteries, a contrast dose reduction from 0.15 to 0.05 mmol/kg did not compromise image quality, acquisition speed, or spatial resolution [5]. Arterial occlusive disease was detected nearly equally between the two readers, with no significant difference in arterial definition scores.

CE-MRA has been established as a non-invasive alternative to conventional angiography in evaluating peripheral vascular disease [10–12] and can be an alternative to CTA for the diagnosis of acute pulmonary embolism [13]. Lower-extremity MRA is typically associated with the highest contrast dose protocols of all MR imaging techniques, often requiring a double-dose (0.2 mmol/kg) or more of Gd-contrast to be administered [14]. It has been shown that the amount of Gd-contrast needed at 3.0T for lower extremity MRA can be reduced up to one-third of that used at 1.5T (i.e. from 0.3 mmol/kg to 0.1 mmol/kg) [15]. The resulting images at lower contrast doses had better arterial definition than high-dose images, presumably due to lower residual background signal from the initial contrast injection and less venous contamination [16].

Renal CE-MRA quality at 3.0T has also been evaluated with low-dose Gd. Attenberger et al. demonstrated equal image quality for evaluation of the renal arteries comparing 0.1 mmol/kg of gadobenate dimeglumine at 3.0T with 0.2 mmol/kg of gadobutrol at 1.5T [17]. Kramer et al. compared low-dose (0.1 mmol/kg) gadopentetate dimeglumine at 3.0T with conventional digital subtraction angiography (DSA) for evaluation of renal artery stenosis in 29 patients, yielding good to excellent quality images with sensitivity and specificity of 94% and 96% respectively [4]. These findings suggest that at 3.0T, the contrast dose in current practice is likely higher than needed, and can be lowered without negatively impacting spatial resolution or overall image quality.

Current CE-MRA techniques using conventional Gd-contrast agents are limited by the need to acquire images relatively quickly during the first pass of contrast material through the vessels of interest. Newer, intravascular Gd-based contrast agents can help overcome these limitations. Gadofosveset trisodium, a protein-binding intravascular contrast agent that has recently obtained FDA approval for use in CE-MRA of the aorto-iliac segments, differs from other gadolinium-based contrast media by having a considerably longer intravascular lifetime and higher relaxivity [18]. Gadofosveset requires smaller total amounts of contrast (Figure 3) and extends the imaging windows up to 60 minutes or more. Images can then be obtained during the steady-state phase after the administration of IV contrast, permitting longer scan times to acquire very high spatial resolution CE-MRA images. A study by Klessen et al. [18] demonstrated that 10 mL of Gadofosveset trisodium produced qualitatively better images with higher arterial contrast compared to 30 mL of gadopentetate dimeglumine. Further optimization of the injection protocol is speculated to further improve the results found in this study.

Parallel imaging at 3.0T

Parallel imaging further enhances the benefits of 3.0T by undersampling the area of interest as a tradeoff for increased image acquisition speed. Parallel imaging has been applied to CE-MRA to reduce scan time and improve spatial resolution by improving anatomic coverage and removing aliasing artifact with the use of multiple channel coils (Figure 4) [19]. The individual coils, which have varying spatial sensitivities, are used to simultaneously receive MR signal following a single radiofrequency (RF) pulse. This permits faster image acquisition with fewer motion artifacts, less RF excitations pulses, and lower energy burden for the patient, but some aliasing is present due to missing k-space data as a result of undersampling. In a study by Fenchel et al. [20], high-quality CE-MRA with integrated parallel acquisition technique (iPAT2) and single contrast injection has been shown to produce adequate image quality of the entire arterial vasculature with acceptable SNR and CNR values for whole-body applications, in less than 60 seconds. Parallel imaging can also increase the anatomical coverage. Lum, et al. [21] recently demonstrated the use of a two-dimensional autocalibrating parallel imaging technique (2D-ARC) to increase the coverage for abdominal CE-MRA. Subjective image quality and vessel conspicuity were graded for healthy volunteers and patients with suspected renovascular disease for MRA with and without 2D-ARC. The results demonstrated equivalent image quality in both methods, with the benefit of a 3.5-fold increase in imaging volume and complete abdominal coverage within the same acquisition time for 2D-ARC MRA. This same technique can also be used to perform high resolution, whole chest MRA in a shorter time, which is important in the evaluation of patients suspected of having pulmonary embolism or who are short of breath (Figures 5, 6).

Limitations and safety concerns for CE-MRA at 3.0T

The stronger magnetic field at 3.0T results in significant challenges and limitations that are yet to be fully overcome. Constructive and destructive interference due to RF field inhomogeneity and increased Specific Absorption Rate (SAR) are major concerns when imaging at 3.0T.

RF field inhomogeneity can result in areas of interference and loss of complete anatomic coverage within the image field. At 3.0T, the resonance frequency of protons in water is 128 MHz, double the value in a 1.5T system, which means that the radiofrequency wavelength is halved from 52 cm to 26 cm. This shortened wavelength can span the dimensions of the field of view for abdominal and pelvic imaging, occurring more frequently in persons with a large body habitus [22]. As two RF waves overlap in the imaging field, constructive or destructive interference can result in areas of brightening or darkening respectively. A similar artifact can occur in persons with large amount of fluid in their abdomen (eg. ascites or pregnancy). Electrical current circulates within the fluid under the strong magnetic field and interferes with the RF field pulses resulting in interference [23]. Advances in coil design, such as multicoil transmit body coils, can suppress eddy currents and improve RF field homogeniety at higher field strengths [24]. In addition to improved coil design, new pulse sequences such as three-dimensional tailored RF pulses have been shown to improve homogeneity of the radiofrequency excitation[25].

RF pulses transfer energy to protons within the patient and ultimately generate heat as a byproduct of energy release. Heat produced within the patient can have detrimental physiologic effects and is carefully monitored within the imaging setting, with current limits of total body heating set by the FDA at 4 W/kg for the whole body over a 15 minute period [26, 27]. SAR provides an estimate for the energy deposited in the tissue by the RF pulse and increases with the square of the resonance frequency. At 3.0T, the resonance frequency is double that of a 1.5T system, and thus the SAR is increased fourfold [2]. Modified pulse sequences, acquisition techniques, and hardware designs are being developed to aid in management of the increased SAR at higher fields. The use of parallel imaging also provides an important solution to this problem, as the multiple detector coils used to simultaneously encode a larger anatomic region serve to both decrease acquisition time and decrease the number of RF pulses needed to acquire an image.

Time-of-Flight MRA

Time-of-flight (TOF) is the most commonly used NCE MRA technique, especially for peripheral and intracranial applications. TOF relies on the suppression of the background signal by rapid slice-selective radiofrequency excitation pulses that saturate the signal from stationary tissue, resulting in suppressed background signal [30, 46]. Because the venous signal could potentially obscure the visualization of the adjacent arteries, the venous flow is usually selectively suppressed by applying a saturation band on the venous side of the imaging slice to null the signal as it enters the slice being imaged. This same principle can be applied to the diaphragm during respiration and the heart during the cardiac cycle. In tissue planes with high flow velocity, the incoming blood will be free of the excitation pulse that saturates the background tissues resulting in strong signal intensity. Slow blood flow or stasis, retrograde filling, tortuous vessels, or vessels in the same plane as the image slice result in saturation of the blood flow in the image volume and poor vessel visualization.

TOF acquisitions can be performed using 2D or 3D sampling, with 3D TOF being most commonly used for intracranial vasculature due to the tortuous nature of the arterial tree, tendency for flow within the imaging plane, and need for high spatial resolution [46]. 2D TOF angiography is used more often clinically in the evaluation of the carotid arteries (Figure 8) and peripheral vasculature (Figure 9), which is oriented orthogonal to the imaging plane [47]. While the saturation of protons within the in-plane vessels is the greatest limitation of TOF, it can be overcome by the use of progressively increasing flip angles through the slab to compensate for the saturation of blood flowing into the slab [48], multiple overlapping thin slab acquisition (MOTSA), which acquires the image volume as multiple thin 3D slabs and has less signal saturation than in a single-volume 3D acquisition [49].

ECG-gating has been successfully applied to CE-MRA techniques in the thoracic aorta, where cardiac motion can result in blurring of the vessel wall in the ascending portion of the aorta [50]. For imaging the peripheral arteries, where blood flow depends on the phase of the cardiac cycle, systolic gating can be used to time the image acquisition during peak blood flow [30]. Lanzman et al. [51]recently describe the use of a promising novel ECG-gated 3D NCE-MRA technique in patients with peripheral artery disease, showing adequate image quality and disclosure of significant arterial stenoses in the lower extremities without the need for exogenous contrast media.

Steady-State Free Precession MRA

Balanced steady-state free precession (SSFP) techniques are popular for NCE MRA because image contrast is determined by T2/T1 ratios, which leads to inherently bright blood images with little dependence upon blood inflow [30]. Both arteries and veins have bright signal with SSFP MRA, which makes this technique well suited for thoracic MRA applications (Figure 10) where the vessels are larger and where evaluation of both arterial and venous structures is important (i.e. in congenital heart disease). In clinical scenarios where venous signal may interfere with the interpretation of the MRA (i.e. renal MRA), venous inflow suppression techniques can be applied to SSFP MRA techniques to obtain purely arterial MRA images.

In a retrospective analysis by François et al. [52] of 23 patients who underwent both CE-MRA and 3D SSFP of the thoracic aorta, measurement of the aortic diameter was essentially equal between the two methods with notably superior visualization of the aortic root using 3D SSFP. A separate study compared CE-MRA to 3D SSFP for the evaluation of pulmonary veins (PV) prior to radiofrequency ablation surgery, and the 3D SSFP images demonstrated accurate PV diameter measurements with superior SNR and CNR [53]. A study by Krishnam et al. [54] demonstrated that free-breathing ECG-gated SSFP MRA of the thoracic aorta had equal diagnostic sensitivity and specificity compared to CE-MRA in 50 patients with suspected thoracic aorta disease. Independent qualitative and quantitative image analysis showed both techniques providing excellent visibility grades of all aortic segments. SSFP MRA demonstrated better visibility of the aortic root and had higher SNR and CNR values for all segments, while allowing the patient to breathe freely during imaging.

3D SSFP MRA has also been applied to the evaluation of the renal arteries. Maki, et al. [55]compared 3D SSFP MRA to CE-MRA at 1.5T in 40 patients and showed that 3D SSFP MRA had a sensitivity of 100% and specificity of 84%. Similarly, Wyttenbach, et al. [56] evaluated 53 patients suspected of renal artery stenosis with 3D SSFP and CE-MRA at 1.5T, with 3D SSFP MRA having a sensitivity and specificty of 100% and 84%, respectively. A study by Lanzman et al. [57] compared the image quality and visibility of renal arteries at 1.5T and 3.0T and demonstrated a significant gain in SNR and CNR at 3.0T of 13-16% and 16-23% respectively, with the greatest improvement of mean image quality at the segmental artery branches. The gain, while significant, is less than expected by the theoretically doubling of SNR anticipated at 3.0T due to SSFP relying on contrast from T2/T1 ratio.

Arterial spin labeling (ASL) is a technique that can be combined with SSFP to enhance image quality through improved background tissue suppression. Protons upstream of the imaging field are "tagged" with an inversion pulse to provide contrast. Background tissue can be suppressed by subtracting the untagged image from the tagged blood image in two acquisitions [58] or by applying a spatially nonselective tag pulse of the entire imaging field in addition to the tag pulse applied to the arteries of interest in a single acquisition [30]. ASL with SSFP provides bright-blood, venous-free images with high SNR especially suited for carotid and renal artery imaging (Figure 11) due to decreased sensitivity to flow artifacts [30]. The complex vasculature of the aorta relative to the renal arteries is well visualized in this technique, and initial clinical experience has shown comparable results to CE-MRA in both healthy volunteers and patients with renal artery stenosis (Figure 12) [59, 60]. Using this type of sequence in 67 patients suspectec of renal artery stenosis, Glockner et al. [61] found that SSFP provided diagnostic images in most cases, but having a higher incidence of false positive and negative results compared to CE-MRA.

ASL is limited by relying on arterial velocity to replace blood in the imaging plane with tagged blood. In peripheral arteries with slower flow, the inflow of tagged blood can approach the T1 of the surrounding tissues, thus eliminating the tagging effect. This can be partially overcome by the multiple, thinner-slab acquisitions, but at the expense of longer imaging times.

Phase-Contrast MRA

Phase-Contrast (PC) MRA generates an image by applying a bipolar velocity-encoding gradient during the pulse sequence twice in opposing directions, which results in a net phase change of zero in stationary tissues while applying a phase change in moving blood, producing a signal. Signal strength is proportional to the velocity of moving blood, and the strength of the bipolar flow encoding gradient, which is prescribed by setting the Velocity Encoding (Venc) value. The Venc describes the maximum velocity that can be accurately encoded without aliasing, similar to Doppler velocity measurement. Thus, phase-contrast MRA provides anatomic images of vessels, in addition to hemodynamic data, about flow, unlike TOF and CE-MRA techniques. The intravascular signal loss on 3D PC MRA at and distal to a hemodynamically significant stenosis (Figure 13) is due to intravoxel phase dispersion related to turbulent flow, and can be used to estimate the hemodynamic significance of stenoses [62]. PC MRA can be used to identify the direction and velocity of flow, and has better background suppression compared to TOF. Its use is limited by longer image acquisition times and higher sensitivity to changes in velocity and magnitude of blood flow during the cardiac cycle [63]. At 3.0T, although there is not increased accuracy of flow measurements compared to 1.5T, there is greater signal and less noise measured for a given VENC. This allows VENC to be increased, reducing aliasing artifacts in regions of higher flow without increasing overall image noise to unacceptable levels [64].

Traditionally, PC MRA was performed with three-directional velocity encoding without any temporal information to obtain a "complex-difference" MR angiogram. With this approach, each acquisition was repeated three times with a different velocity-compensation direction and once without flow compensation. Because four acquisitions are needed for 3D PC MRA, the scan times are long and the imaging volume is limited. Parallel imaging techniques [65] and 3D radial undersampling [66], or Vastly undersampled Isotropic Projection Reconstruction (VIPR), have been used to reduce scan time without compromising coverage or resolution. In addition, these image acquisition acceleration techniques have enabled the acquisition of temporal information in addition to the standard 3D PC MRA acquisition, resulting in four-dimensional (4D = three-dimensional spatial encoding, three-directional velocity encoding, and time) PC MRA for a variety of vascular applications. While these newer 4D PC MRA sequences can be used for NCE MRA, the future direction of these techniques lies in the additional hemodynamic information provided. In contrast to conventional two-dimensional (2D) PC MRA, where the vessel of interest must be known prior to scanning and the image plane must be prescribed at the scanner during the examination, 4D PC MRA techniques permit the post priori evaluation of flow velocities of any vessel within the imaging volume from the same acquisition. In addition, the 4D PC MRA techniques can be used to qualitatively evaluate the complex flow patterns within the cardiovascular system (Figure 14) [65, 67–69] and calculate various hemodynamic parameters non-invasively, including pressure gradients [70, 71, 61], wall shear stress, and oscillatory stress index [68, 72]. Implementation of these techniques into clinical routine is currently limited by our ability to process and interpret the large amount of data generated by these sequences.