Abstract

William G. Bradley
Department of Radiology, University of California, San Diego, USA

Introduction
Over the past two decades MRI has been used clinically, progress has been more sporadic than steady. If one were to plot it out, MR advances would be more of a series of steep climbs followed by plateaus than a slow steady climb. The steep parts in the past were the transition from resistive to superconducting magnets in the early 1980’s, the transition from low field superconducting to the current 1.5T superconducting systems in the mid 1980’s, the introduction of echo planar (EPI) systems in the mid 1990’s and the introduction of “cardiovascular systems” with even stronger, faster gradients in the late 1990’s.

3T, SENSE and Phased Array Coils
Now, it appears as if we are approaching another steep part of the curve due to the synergy provided by the greater 3T field strength of the latest commercial MRI systems and “parallel imaging”, known more commonly as SENSE (Philips), ASSET (GE), and iPAT (Siemens). (With apologies to GE and Siemens, this technique will henceforth be called SENSE, since Philips invented it). SENSE is a technique that trades signal-to-noise (S/N) for speed. Just as S/N is reduced by (40%) whenever the number of excitations is halved in a conventional spin echo, the same is true of SENSE: a SENSE factor of 2 halves the acquisition time and reduces the S/N by 40%. So SENSE is most useful when there is excess S/N, eg, at 3T.

SENSE requires the use of phased array coils. The number of coil elements determines the theoretical limit on the SENSE factor. The name SENSE comes from SENSitivity Encoding. SENSE works by intentionally reducing the field of view (FOV) and the number of phase encoding steps which in turn reduces the acquisition time. This would normally lead to wraparound artifact or “aliasing”. However, by knowing the local sensitivity of each coil in the phased array, the aliasing can be “unwrapped”. Phased array coils also have the advantage of higher S/N than standard quadrature RF coils. Thus the greater S/N of these coils and the higher field strength can be traded for speed or higher spatial resolution at the same acquisition time.

A major beneficiary of 3T and SENSE technology will be MR of the breast. At 1.5T, typical spatial resolution for a single breast exam is 1mm in plane with a slice thickness of 2.5-3 mm. At 3T with SENSE, spatial resolution of 0.5 mm can be obtained with a slice thickness of 1 mm. This added spatial resolution allows much better evaluation of lesion borders to depict spiculation. MR of the breast is already essentially 100% sensitive for invasive breast cancer. Combining the kinetic data with this improved spatial resolution will undoubtedly improve the specificity of MRI for detection of breast cancer as well.

Phased array coils offer an additional advantage in the brain at 3T. Using a standard birdcage head coil, the center of the brain tends to be brighter due to dielectric effects. Phased array coils at 1.5T tend to have greater signal immediately beneath the coil elements at the periphery of the brain. Using phased array coils at 3T, the signal throughout the brain evens out.

1024 MRA of the Circle of Willis
1024 MRA of the Circle of Willis (COW) has been a sort of Holy Grail for neuroradiologists. Earlier MR systems either did not have the computer power to acquire a 1024x1024 image or they didn’t have the necessary S/N. Attempts to boost S/N with gadolinium were complicated by venous enhancement. Since the acquisition time for the gradient echo techniques used for 3D time of flight (TOF) MRA is TR x Np x Ns, increasing the number of phase encode steps (Np) to 1024 would either lead to an excessively long acquisition time or mandate a reduction in TR. At some point, a very low TR limits the time available for inflow of unsaturated blood and reduces flow related enhancement which is the basis for TOF MRA. Increasing the matrix size from 512 to 1024 along one axis decreases S/N by 50% (assuming FOV is held constant). Halving pixel dimension along both the phase and frequency axes reduces S/N to 25% of what it was (holding acquisition time and slice thickness constant). Since the eye (like most organs) responds logarithmically to physiologic stimuli, a loss of 75% of the original S/N will be most noticeable at low S/N levels. The added S/N afforded by 3T magnets and the latest generation of 8 channel head coils, allows this drop to go essentially unnoticed. With a rectangular FOV, a smaller number of phase steps can be used to achieve the same spatial resolution, reducing the acquisition time.

A 1024 x 608 acquisition over a 16 x 12 cm FOV, yields pixels 160 x 200 microns. This compares to the typical 1024 digital subtraction angiogram (DSA) which has an in plane spatial resolution of 250 microns and typical CTA which has spatial resolution on the order of 500 microns. The slice thickness of the MRA is 0.8mm (800 microns) which is zero interpolated to 400 microns. With a SENSE factor of 2, the acquisition time is on the order of 7 minutes.

On a 1024 MRA, all the arteries of the COW appear larger than in lower resolution MRAs. This is due to the fact that intravoxel dephasing has been minimized at the periphery of the vessel where the velocity changes (and the phase dispersion) are greatest. The initial results are impressive. For the first time, the ophthalmic, anterior choroidal, and lenticulostriate arteries can be visualized over several centimeters – and the results are only going to get better. This level of detail will allow diagnosis of spasm from SAH, vasculitis, and intracranial atherosclerosis, greatly extending the clinical applications of MRA in the brain. Although the spatial resolution of MRA can now be better than that of DSA, detection of abnormalities is a function not only of spatial resolution but also of signal to noise (remember the contrast detail curves and modulation transfer functions from your residency?). DSA also has the advantage of temporal information. In order to use a technique like TRICKS to get temporal phases from an MRA, contrast would have to be used and 1024 spatial resolution would probably need to be reduced to get clinically reasonable spatial resolution.

As a result of these examples, the improved spectral separation for MR spectroscopy, and quite a few additional improvements offered by the combination of higher fields, SENSE, and phased array coils, I would predict that 3T will replace 1.5T for premium commercial systems within the next 5 years.

MR for Acute Stroke Evaluation
The decision to administer tPA to patients with acute stroke symptoms is currently based on unenhanced CT. If there is hemorrhage or hypodensity in more than one third of the MCA territory, tPA should not be given. Despite the fact that only 2% of all stroke patients get to the ED in time to have contrast, more than half of those that do get it probably shouldn’t. There are several reasons for this. For a start, up to 20% of patients with a clinical diagnosis of stroke in the ED have another diagnosis, eg, complicated migraine with hemiplegia or hemisensory deficits, Todd’s paralysis following a seizure, or inflammatory processes like cerebritis. 10-20% of patients have additional clinically silent strokes which are more than 3 hours old and would thus be a contraindication for tPA. (10% of patients bled outside the area of primary symptomatic infarction in the original NINDS stroke trial, presumably due to the presence of embolic disease and undetected older infarcts). Finally, 20% of patients who get tPA don’t have an ischemic penumbra, ie, brain at risk for extension of the infarct. These patients probably shouldn’t receive tPA.

I would predict that over the next decade MRI will gradually replace CT as the primary imaging modality for stroke triage. For a start, MRI can do what CT does, ie, detect hemorrhage. Gradient echo images are more sensitive than CT for acute hemorrhage and the b=0 image from the diffusion study (a spin echo EPI image) and the baseline images from the perfusion data set (gradient echo EPI image) are even more sensitive than traditional gradient echo for detecting acute parenchymal hemorrhage. FLAIR has been shown to be 100% sensitive for the detection of subarachnoid hemorrhage detected by CT. In fact, one study demonstrated that FLAIR was not only as sensitive as CT: it was 100 times more sensitive!

MRI can also identify the patients who shouldn’t be getting tPA. Diffusion and perfusion imaging can confirm the presence of ischemia. Diffusion imaging can show older, clinically silent infarcts. And if the diffusion and perfusion studies show lesions of the same size, there is no need for tPA since there is no ischemic penumbra.

Intraoperative MRI
Intraoperative MRI will increase as primary guidance for delivery of gene and stem cell therapy. Gene therapy is currently administered for Alzheimers and Parkinsons using a stereotactic frame. Following a 1 cm burr hole, CSF leaks out and the brain shifts from its preoperative position. And even if there were no brain shift, the sterotactic coordinates in Talaraich space are based on the brain of a 40 year old, alcoholic French woman. Both effects contribute to the lack of accuracy of blind stereotaxy in any given individual. If the gene therapy is delivered by a viral vector to a point 2 mm beyond the basal nucleus of Maenert, the patient is essentially dead. Thus accurate guidance is of paramount importance. With direct visualization by MRI, the tip of the needle can be well visualized and the position confirmed before the gene therapy is administered. Given the prevalence of Alzheimers and Parkinsons, if new gene therapies are proven to be efficacious, this alone will likely drive the need for intraoperative MRI.

MRI has been used for brain tumor resection for almost a decade. Intraoperative scanning allows the surgeon to proceed carefully, avoiding eloquent structures. It also allows them to get it all out. While this is never truly possible for high grade gliomas, it is for low grade lesions. The prognosis for low grade lesions is inversely related to the amount of tumor left behind as even a small amount of low grade tumor will undergo malignant degeneration and eventually kill the patient. Unfortunately, 80% of the time the neurosurgeon thinks he has achieved a gross total resection, tumor can be found by MRI. Thus MRI is critical for the resection of low grade lesions.

MR-guided focused ultrasound (FUS) has been used to monitor the ablation of breast tumors and uterine fibroids. The advantage of MRI is that it can monitor temperature change and steer the point of confluence of a thousand ultrasound transducers to the remaining portion of the lesion. Using CT to measure the thickness of the skull, researchers at Brigham and Women’s hospital are already using MR-guided FUS to ablate malignant brain tumors. At UCSD, we are planning to use Graeme Bydder’s new ultrashort TE MR techniques to do the same thing with MRI alone. If the ultrasound can be operated in shock wave mode (like lithotripsy), it might even be possible to use such a system to blast a clot in an acute stroke patient.

Neonatal MRI
Many neonates are too fragile to leave the Neonatal Intensive Care Unit (NICU) to have an MRI study. Researchers at UCSD are evaluating a small bore 3T magnet which would site in the NICU and allow the neonate to have both MRI and MR Spectroscopy. The isolette would maintain the proper humidity and temperature while monitoring the baby. Rather than the usual MRI scan which is a 30 minute affair, neonatal MRI might go on continuously, allowing, for example, the evaluation of new neuroprotective agents for the treatment of ischemic-anoxic disease using MR spectroscopy. At 3T, glutamate and glutamine can be distinguished – which should be useful for monitoring treatment of ischemic lesions as well.

Targeted Receptor Contrast Agents
One of the goals of molecular medicine and molecular imaging is the development of novel contrast agents which are targeted to specific receptors. While PET or SPECT will likely be the first imaging techniques to be used for such agents, MRI is likely to follow quickly if the agent can be tagged with enough gadolinium. Such agents are likely to target the proteins which are elaborated by upregulated genes to demonstrate, for example, angiogenesis.

Atherosclerotic Plaque Characterization
For the past 75 years, radiologists have focused on narrowed arteries as a sign of atherosclerosis. We have focused on the lumen rather than on the plaque which was causing the luminal narrowing. Contrast-enhanced MRA of the carotids has already replaced catheter angiography in most institutions. The coronaries are likely to take a few years more.

The current popularity of coronary calcium scoring is likely to be supplanted by CTA of the coronaries and, over the next 5 years, by MRA. While the lumenography of CTA may actually still be better than that of MRA, it is anticipated that plaque characterization by MR is likely to be superior.

Plaque characterization involves differentiating “vulnerable” from “stable” plaque. Vulnerable plaque is more likely to rupture, leading to acute occlusion of the vessel and then to myocardial or cerebral infarction. It is characterized by a lipid core full of inflammatory cells and oxidized LDL with a thin fibrous cap. Stable plaque has a thicker fibrous cap and less lipid. While CT can detect calcification (a marker of plaque stability), MRI can show the size of the lipid core (bright on T2WI) and the thickness of the fibrous cap (enhancing on T1WI). Antibodies to oxidized LDL have already been developed by researchers at UCSD and contrast agents to target vulnerable plaque are under development.

Conclusion
To get an idea of where MRI will be in the next 10 years, look at the last 10 years. We can expect increasing applications as a result of higher fields, faster, stronger gradients and multichannel RF subsystems to take advantage of techniques like SENSE. While PET-CT is currently the rage, it wouldn’t surprise me if PET-MR systems become available in the future. Research techniques like diffusion tensor and fMRI will become commonplace in clinical application. To complement fMRI, MR will almost certainly be linked up in some fashion with MEG (magnetoencephalography), to produce magnetic source images – although the linkage is likely to be more through image fusion than combined units. One thing is certain: MRI will continue to evolve and will remain the dominant imaging technique in medicine.