A fundamental limitation that current MRI techniques suffer from can be thought of as simply too low of a ratio of signal to noise. As users look for ways to get more signal, an increasing number are using higher field strengths. At MGH-NMR Center, Massachusetts General Hospital, Boston, we have found significant benefit by moving from 1.5T MRI to 3T MRI in a variety of routine and investigative settings.
the currency of MRI
An important concept in considering the benefits of high field imaging is that the signal-to-noise ratio (SNR) can be viewed as a type of currency that the radiologist can choose to spend in a number of ways. Improved SNR might be used to increase the imaging matrix, to choose thinner slices, to reduce the number of signal averages, or, in some cases, to decrease the amount of contrast agent administered. This flexibility of improved SNR is one of its greatest advantages. Once an enhanced SNR is available, how to use it best is up to the radiologist.
It is also important to note that increasing field strength is not the only way to improve the SNR. A fourfold increase in imaging time will lead to a doubling of the SNR. This means that SNR calculations always assume a fixed amount of imaging time, since one can improve SNR simply by imaging longer. SNR can also be improved by using improved receiver coils (such as surface coils, singly or in arrays); in some instances, by shortening echo times (made possible by improved gradient hardware); or by using novel pulse sequences. These sequences often optimize the contrast-to-noise ratio (CNR), but they sometimes boost the SNR as well.
Most new 1.5T systems in use today are already using optimized hardware and software, so these strategies are less applicable. While longer imaging times might provide better SNR, increasing the imaging time by a factor of four is simply not possible in many situations. For example, the examination may already be as long as feasible, the patient’s condition may not be stable, or there may be economic reasons (such as the need to maintain a high rate of patient throughput). Therefore, using a higher field strength is an option that merits exploration.
Areas of Potential Benefit
|Figure 1. Ultra-high resolution T2 weighted images showing the Circle of Willis. Note the high level of detail evident, in particular the lenticulostriate and perforating vessels arising from the middle and posterior cerebral arteries (arrows). While a similar matrix could be used at 1.5T, it would produce images with increased noise (a grainy appearance), and less diagnostic confidence. Courtesy of Larry Wald, PhD; acquired on a Siemens Allegra 3T with a 4 channel array coil. Acquisition parameters: 15 echo Turbo Spin Echo: TR/TE = 4000/90ms, FOV = 156x180mm, 450 X 512, 2mm slice thickness 0.35 x 0.35 x 2.0 mm3 voxels, 9min 16 sec scan time, 15 slices.
Conventional MRI already provides high-quality imaging in many settings. There are, however, some instances in which finer detail would be desirable, particularly if the imaging time need not be increased to obtain it. Figure 1 shows an example of the intracranial vasculature near the circle of Willis, demonstrating small perforating arteries not typically visible. While the increased matrix size would be possible at 1.5T, the resulting voxel size would lead to a grainy image. Such a fine level of detail is appealing to surgeons, radiologists, and radiation therapists, as well as to neuroscientists. Most CT images have a voxel size that is one quarter or less that of a typical MRI image. While MRI provides superior CNR for most lesions, increased spatial resolution would be welcome in most arenas. Specific diseases in which obtaining increased spatial resolution for conventional images might prove to be cost effective include tumors (in which presurgical planning could be assisted) and epilepsy and congenital abnormalities, which are often subtle and difficult to detect.
|Figure 2. Magnetic resonance angiography. Left Image (a) was performed at 1.5T on a Siemens Sonata, image (b) at 3T on a Siemens Allegra. Note the increasing vessel conspicuity and the level of detail. Both were 1024×384 matrix sizes; the Allegra image took approximately 7 minutes; the Sonata approximately 16 minutes.
We have found substantial improvement in the appearance of MR angiography (MRA) images at higher field strengths, and we are currently investigating the specific cause of this improvement. Figure 2 shows 1.5T and 3T images, demonstrating the high level of detail available at 3T. In this case, the matrix sizes are the same; the improvement in SNR shows as a reduction in the graininess of the image and in the increased conspicuousness of the fine vasculature. We anticipate that, with additional work to optimize parameters for 3T scanning, MRA will improve further; it may eventually rival catheter radiographic angiography. Diseases that could benefit from this improvement in spatial resolution include aneurysms and atherosclerotic disease. Improved diagnostic confidence in MRA could preclude more catheter radiographic angiography studies and allow more confident screening of patients with congenital (or other) predispositions to forming aneurysms.
With the recent focus on dynamic contrast-enhanced angiography, the benefit of improved SNR at 3T will be even more apparent. This is because these contrast-enhanced techniques are inherently SNR-limited. They depend on acquiring multiple images during the rapid passage of contrast through the arterial tree, and increasing the imaging time is not an option.
|Figure 3. fMRI at 1.5T and at 3T. These images are computed eccentricity phase maps, used for mapping retinotopy in the visual field. The images are on flattened visual cortex, taken from the calcarine fissure (area circled on small image of brain cortex at top left). The subject is shown expanding rings, and the delay from the onset of the stimulus is mapped as a phase change. Colors represent isophase regions, with the white areas representing one particular phase value. Note that at 3T, the signal strength is higher and allows mapping of a larger area of visual cortex, whereas at 1.5T, many areas are simply not visible, since the noise hides the reaction of the brain tissue. Many more white areas are present at 1.5T, indicating that the ability to distinguish noise from signal is decreased. Courtesy of Anders Dale, PhD, and Bruce Fischl, PhD, MGH-NMR Center.
The area of functional MRI is, perhaps, where high-field magnets will be of greatest impact. Blood-oxygenation-level-dependent imaging gains an additional boost from a higher field because susceptibility effects generally increase as the square of the field strength. Hence, at 3T, the level of signal change can be up to 7%, compared with 1.5% or less at 1.5T. This benefit is particularly apparent in investigating subtle changes and in using paradigms that require extra SNR, such as single-trial designs or event-related studies. One way to measure the apparent benefit of higher field strength is to determine the extent of activation for a given paradigm. With better SNR, additional areas should be resolved from the background noise. Figure 3 demonstrates the results of a visual-stimulation paradigm at 3T and at 1.5T, and indicates that extensive additional information is available at 3T.
Studies designed to investigate dementia, drug addiction, migraine, and other disorders are underway at our institution, as well as at others, and investigators are demonstrating benefits available at 3T.
|Figure 4. Perfusion MRI at 3T. Maps of relative cerebral blood volume (rCBV) and relative cerebral blood flow (rCBF) can be created from dynamic data acquired at 3T. These images demonstrate abnormally elevated rCBF and rCBV in a residual grade 3 tumor. Note the excellent grey / white ratio, consistent with known differences in grey matter and white matter blood flow and blood volume. Image acquired on Siemens Allegra, spin echo EPI.
Diffusion and perfusion MRI are relatively new techniques that typically push SNR requirements to the limit, so they also benefit from higher field strength. Perfusion-weighted imaging has an additional way to spend the increased SNR currency that 3T scanning yields. The susceptibility changes that are the basis of perfusion MRI can be obtained using a lower dose of gadolinium-based contrast agents than is possible at lower field strengths. In effect, the higher field strength means that half the contrast dose can be used to obtain the same effect. Figure 4 shows perfusion MRI images obtained at 3T and indicates that creating maps of relative cerebral blood flow, as well as relative cerebral blood volume, is easily done using the higher SNR of 3T MRI. Perfusion measurements have often been limited by SNR, since the first pass of a contrast agent occurs so quickly and, therefore, there are relatively few data points from which to compute hemodynamic parameters. The additional SNR boost (assuming that the same dose of gadolinium is given) can allow more precise measurement of hemodynamics, particularly in disease states such as cerebrovascular accident (CVA), where hypoperfusion is present and only a small amount of contrast agent might be arriving in a voxel.
Diffusion MRI is also limited by SNR: many sites typically acquire multiple signal averages at 1.5T, but still have insufficient SNR for the full study of phenomena such as anisotropy and behavior of the full diffusion tensor in disease states such as CVA. The additional SNR boost should help these studies as well.
The quest for higher field strength will continue. Nearly 50 3T and higher human-sized MRI systems are in use around the world, with another 40 to 50 planned for installation in the coming 12 to 18 months. A few centers (including ours) are installing even higher field strengths of 7T and beyond. Which of today’s applications (or, perhaps, applications not yet discovered) will provide the most compelling reason to move to higher field strength in the clinical setting remains to be determined, but the movement toward higher field strength for clinical imaging is clearly happening.
Gregory Sorensen, MD, is associate director, MGH-NMR Center, Massachusetts General Hospital, Boston.