Applied Radiology

Dr. DeSena is a partner at Radio-logical Services of New York, Staten Island, NY. Dr. Earls is a partner at Fairfax Radiological Consultants, Fairfax VA.

Contrast-enhanced magnetic resonance angiography (CEMRA) has been rapidly introduced into widespread clinical use over the past several years. Intravenous gadolinium chelates for MR angiography minimize flow artifacts, saturation effects, and long imaging times that degrade time-of-flight (TOF) MR angiography (MRA), and their use in three-dimensional (3D) MRA has produced excellent results. ^1-9 Contrast-enhanced MRA was originally developed and described for use in high-field (1.5 tesla [T]) magnetic resonance imaging (MRI) systems. In routine clinical practice, it is performed almost exclusively at 1.0 to 1.5 T. Low-field MRI systems are limited by slower pulse sequences and lower signal-to-noise ratio (SNR) qualities. This article will describe our initial experience performing CEMRA on a 0.5 T closed-bore system.

Study methods

Gadolinium-enhanced MRA was performed on 8 patients (ages 25 to 78 years; 5 men, 3 women) referred for MRI and MRA evaluation of the aorta (n = 4) or renal arteries (n = 4). Aortic studies were performed to evaluate for aortic aneurysm, and renal studies were performed to evaluate for renal artery patency and stenosis. All examinations were performed on a 0.5 T closed-bore MRI system (Contour, GE Medical Systems, Milwaukee, WI). The 0.5 T system has a maximum gradient strength of 15 millitesla per meter and a maximal slew rate of 17 tesla per meter per second.

Prior to the start of the examination, a 22-gauge intravenous catheter was placed in a forearm vein. All patients were given breath-holding instruction prior to the study and were able to hold their breath for at least 20 seconds. A fast 3D vascular TOF gradient-echo (3D FSPGR) pulse sequence (10 to 13 msec/1.8 to 4.2 msec [repetition time/echo time], 35š to 40š flip angle, bandwidth 31.2 kHz, 3 to 6 mm slice thickness, 36 to 45 * 27 to 33 cm field of view, 256 * 128 matrix, 24 slices acquired in a 30- to 35-second breath hold) was used for the CEMRA studies. The sequence parameters were tailored individually for each case to balance the need for complete anatomic coverage with minimizing overall acquisition time. The 3D FSPGR was a commercially available sequence that provides the low repetition and echo times necessary for breath hold contrast-enhanced vascular imaging. For imaging the aorta, the authors used the standard built-in body coil. For imaging the renal arteries, a wrap-around abdominal flex coil (GE Medical Systems, Milwaukee, WI) was employed for increased SNR characteristics.

Following a fast gradient-echo axial scout series, the anatomic area of interest was selected and a 3D slab was prescribed (in the coronal plane for renal studies and in the sagittal plane for aortic studies). The scan delay time varied from approximately 20 to 30 seconds based on the patient's circulation time as estimated by the attending radiologist (SD). Two complete breath-held sequential data sets were acquired following intravenous injection of gadolinium-DTPA (Magnevist, Berlex Laboratories, Wayne, NJ) at 0.2 mmol/kg into an antecubital vein. Gadolinium-DTPA was injected by hand at a rate of approximately 2 mL/sec followed by a 10-mL saline flush. A power injector was not available. Scan time was approximately 30 seconds for each acquisition. The second acquisition was obtained 10 seconds after the first acquisition to obtain an angiographic data set in the venous phase and to provide a backup acquisition in case of unanticipated prolonged circulation of contrast. After the 3D FSPGR data sets were acquired, they were transferred to an independent workstation (Advantage Windows version 3.1P, GE Medical Systems, Milwaukee, WI). The authors performed multiplanar volume rendering (MPVR), maximum intensity projection (MIP), and 3D volume rendering (3DVR) reformatting to visualize the relevant vasculature.

Data analysis

Signal intensity (SI) measurements and standard deviations were determined by drawing the largest possible user-defined regions of interest (ROIs) over three separate areas of the arterial anatomy (ascending aorta, aortic arch, and descending aorta for aortic exams, and abdominal aorta and each renal artery for renal exams). In addition, ROIs were placed over areas of fat and muscle for image contrast calculations, as well as outside the body in the phase-encoding direction to measure the mean and standard deviation of the background noise. Signal-to-noise ratio of the aorta was defined as the mean SI divided by the standard deviation of the background noise. Aorta-fat contrast-to-noise (C/N) was defined as the SI of the artery minus the SI of the brightest fat within the imaging volume, divided by the standard deviation of the background noise. Aorta-muscle C/N was defined as the SI of the aorta minus the SI of skeletal muscle divided by the standard deviation of the background noise. For each patient, data from the first of two post-gadolinium acquisitions was used since the second acquisition was technically inferior because it was obtained in the venous phase.

Results

The quantitative signal characteristics are listed in Table 1. The SNR and contrast-to-noise ratio (CNR) values compared favorably with those published in similar studies performed at 1.5 T. ⁴ Quantitative data was agreeable for both renal and aortic exams (Table 1). No difference was noted between the mean aortic and renal SNR and CNR values. Signal did tend to decrease at the edges of the magnetic field (Figures 1 and 2). Because of the lower matrix (128 phase encodes) and thicker partitions (3 to 6 mm), the overall resolution was lower than CEMRAs performed on most systems at 1.5 T. This was especially true when the 3D images were viewed orthogonal to the slice selection direction, and when imaging the renal arteries (Figure 3). Each of the eight studies yielded acceptable angiograms and the reader's self-assessed diagnostic confidence was considered high. There was no difficulty depicting the type B aortic dissection (Figure 4) even with a slice thickness of 6 mm, and complete anatomic coverage of the aorta was easily obtained (Figure 5).

Discussion

The advantages and disadvantages of low-field MRI systems have been discussed in prior literature and texts. ¹⁰ The current literature describes a variety of applications of low-field MRI, primarily for musculoskeletal and neuroradiology applications. ^11-13 To the authors' knowledge, however, there is no literature on the use of low-field systems for contrast-enhanced MR angiography. This is unfortunate since low-field MRA may be the only noninvasive means of obtaining angiographic information on patients who may not have access to a high-field system.

Although low-field systems obey the same physical laws as their high-field counterparts, their lower field strength results in noticeable differences between the two platforms. Decreased gradient system requirements allow cost savings in gradient system design at the expense of lower, slower receive bandwidths. Decreased T1 relaxation times permit shorter TR and scan times. Low-field systems have less chemical shift and susceptibility and fewer flow artifacts than their high-field counterparts. Additionally, the smaller fringe field permits smaller site requirements and less-restraining "open bore" configurations. Unfortunately, low-field MRI systems are also subject to numerous disadvantages. Signal-to-noise ratio decreases proportionally with field size. Susceptibility artifact also decreases, and while this is at times advantageous there is decreased sensitivity for detecting calcification, iron, and hemorrhage. There is also a smaller selection of pulse sequences and imaging options because there are fewer low-field systems in the marketplace.

The radiologist must be aware of the limitations imposed upon scan quality by the low-field environment. By far, the most important limitation of low-field systems is their inherently lower SNR and resolution as compared with their high-field counterparts. Selecting the appropriate scan parameters to maximize the SNR of the 3D imaging volume is critical to the success of the low-field MRA examination. In any MRI acquisition, slice thickness is directly proportional to SNR and affects resolution as well as extent of anatomic coverage. We wish to keep slice thickness as large as possible to maximize signal and anatomic coverage, but we must also keep it small to maintain anatomic resolution. When imaging a tortuous aorta, this becomes a problem that must be addressed by adjusting both slice thickness and the number of locations per slab. We routinely use a 3- to 6-mm slice thickness and adjust our number of locations per slab to cover the anatomy accordingly. When imaging renal arteries, we try to use a slice thickness closer to 3 mm to prevent loss of resolution, usually at the expense of time, anatomic coverage, and SNR.

The relatively slower pulse sequences of our low-field system can use only 20 to 24 locations per 3D slab to keep the acquisition time around a reasonable breath-hold time of ¾ 30 seconds. Frequently 30 seconds is the upper limit of time that many older patients can hold their breath. For elderly or respiratory-compromised patients, this may still be too long for a breath-hold, and fewer locations must be chosen with sacrifice of some anatomic coverage. Given a 3- to 6-mm slice thickness and 20 to 24 locations per 3D slab, we can cover a volume of 6 to 14 cm in depth, which we feel is adequate coverage for the majority of patients we encounter.

Low-field systems have less stringent gradient requirements than their high-field counterparts and, as a result, have lower receive bandwidths, which accounts in part for their longer scan times. While this is advantageous for SNR, it imposes a limit on the shortest scan time for a 3D acquisition, and therefore on the shortest practical time for a breath-hold acquisition. In practice we use the highest receive bandwidth available on the system (32 kHz) to obtain a scan prescription that will cover all relevant anatomy in a single breath-hold.

There are several limitations and weaknesses in this report. We do not have conventional angiographic correlation for our studies. A direct comparison with conventional angiography would be very helpful to determine the sensitivity and specificity of low-field CEMRA. We did not prospectively estimate the patients' circulation time by performing a "timing run," as is now commonly performed. ⁴ No automatic triggering or fluoroscopic triggering is available at 0.5 T. The "best guess" technique of determining the scan delay time used here can result in a substantial decrease in image quality if the circulation time is misjudged. Finally, the resolution of these exams is clearly lower than at 1.5 T, and a direct comparison of studies performed at both field strengths would be of great interest.

The limited resolution characteristic of low-field imaging is particularly important when imaging the renal arteries, where vessel diameters of 6 mm are common. This limits low-field MR angiographic imaging of these vessels to assessing vessel patency, course, and congenital anomalies. Grading renal artery stenosis with low-field MR angiography is admittedly difficult, and great care must be taken to use thin slice thickness, proper slab selection, and meticulous bolus contrast administration.

Conclusion

The authors were pleasantly surprised by the image quality of our CEMRA studies performed at 0.5 T, despite the limitations mentioned above. By carefully balancing the imaging parameters, adequate breath-hold CEMRA studies of the aorta and renal arteries can be performed at 0.5 T. Additional evaluation at this field strength needs to be performed to
determine the accuracy of these studies. Institutions that have 0.5 T systems can consider using them to perform this technique when a high-field system is not available and the patient's clinical condition prohibits transfer to another facility. AR