The MRI main field strenth has been constantly increased over the past decades. Today, scanners with 3T, 7T and even beyond are in use, however ultra-high field (≥7T) systems are mostly used in research centers although the transition into hospitals is expected. The reasons for using (ultra-)high fields are multifold and will be outlined in this presentation. Along with these benefits go a larger range of challenges, which are among the reasons for the rather slow transition of UHF into clinical applications. Solutions to most of these challenges will be presented and applications will be highlighted.
Since the beginning of magnetic resonance imaging, the main magnetic field strength (B0) has constantly increased. Currently, most clinical MRI systems operate at B0 of 1.5 Tesla (standard field strength) or 3 Tesla (high field strength). In 1998 an 8T system was installed in Ohio and a year later the first 7T came up to field in Minnesota. Since then, the number of ultra-high field (UHF) systems (B0≥7T) has substantially increased to approximately 60 systems worldwide today.
Most of the UHF systems are being used for research purposes; however, the transition into hospitals for dedicated applications is expected. The reasons for using high and ultra-high field compared to standard field strength are multifold and will be outlined in the following. Along with these benefits go a larger range of challenges, which are among the reasons for the rather slow transition of UHF into clinical routine.
One of the major advantages of using higher main field strength is an increase in signal-to-noise ratio (SNR) (1, 2), which is an important quantity that impacts the maximum spatial and temporal resolution and the ability to depict structures in MR images. While the magnetization increases linearly with B0, the theoretical SNR gain is more difficult to assess (3). In practice, an linear to quadratic increase in SNR with field strength has been reported, however, the SNR at UHF is spatially dependent (1, 2, 4). An SNR increase by a factor 4-6 from 1.5T to 7T appears to be realistic. The higher SNR is often used to increase the spatial or temporal resolution. Higher spatiotemporal resolution, however, requires an increased measurement time, thus acceleration techniques become increasingly important with rising field strength. Fortunately, the parallel imaging performance increases with B0, allowing for higher acceleration factors (5). Other acceleration methods, such as KT-acceleration, compressed sensing or Multiband imaging are promising with regard to limiting scan times (6-9). The increased SNR provided by UHF is also promising for nuclei other than protons, such as 13C, 17O, 19F, 23Na or 31P. Signals caused by those “X-nuclei” are typically orders of magnitude lower than for protons, because the concentration in the human body and the gyromagnetic ratio γ is lower compared to protons (note that the MR signal scales with γ3). The higher SNR allows for acceptable resolutions within clinically feasible scan times, which may allow to investigate physiological processes in patients.
The T1 relaxation constant increases with B0 field strength (10). For example, T1 values of white matter amount to 656ms at 1.5T, 838ms at 3T and 1220 at 7T. A longer T1 value in combination with higher SNR is particularly beneficial for inflow-based MR angiography techniques, such as time-of-flight (TOF) (11, 12). In this case the contrast benefits twice from higher B0: the longer T1 values result in stronger suppression of the static background tissue while the full magnetization of the freshly inflowing blood provides stronger signal compared to lower fields. Blood flow quantification techniques such as 4D flow MRI similarly benefit from higher fields (13). For other sequences, however, a longer T1 constant may need to be compensated by longer TR in order to achieve the same contrast. While T1 increases with static field strength the T2* relaxation constant decreases. In practice, this is typically rather challenging, as it results in signal loss and requires shorter echo times to compensate this effect.
The magnetic susceptibility χ of the tissue causes small deviations ΔB0 of the local magnetic field. Because ΔB0 scales linearly with B0, susceptibility induced changes in ΔB0 at tissues will increase at UHF, thus susceptibility weighted imaging (SWI) or quantitative susceptibility mapping (QSM) benefit from UHF (14, 15). The sensitivity towards the susceptibility of deoxygenated blood also increases, thus blood-oxygenation level dependent (BOLD) contrast used in functional MRI increases with B0 (2).
MR spectroscopy also benefits substantially from higher magnetic fields due to an increased chemical shift (16). Thus, the frequency difference between different metabolite resonances increases linearly with B0 allowing for higher spectral resolution and better spectral quantification.
In order to excite the spins a radiofrequency (RF) field is applied by the RF coil, consisting of a transmit magnetic (B1+) and an electric component (E). The frequency of the RF field increases linearly with B0 and at 7T the resulting RF wavelength achieves values of about 11cm in the tissue. Therefore, at 7T the wavelength is in the order of organ sizes, which results in spatially heterogeneous B1+ amplitudes and phases at UHF (17) causing spatially dependent flip angles and unwanted spatially varying contrast. In practice, most commercially available head coils generate a birdcage or circularly polarized (CP) mode during excitation that leads to high FA in the brain center and reduced FA in the brain periphery. Typical values in the periphery are about 50% of the center B1+ amplitude values at 7 Tesla, however, even complete FA voids can occur depending on the RF coil and field strength.
Another substantial challenge is related to the power of the applied RF field. In a simple model, a quadratic increase of the RF power with field strength is expected while a more detailed analysis shows deviations from the quadratic increase towards slightly lower values starting at around 3T (3). Higher RF power has substantial impact on MR hardware and MR safety. The required peak RF voltages increase and therefore stronger amplifiers are required. However, a more important issue is given by the fact that the specific absorption rate (SAR), i.e. the RF power deposited in the tissue increases. With increasing B0 the focus shifts from the global SAR, i.e. the SAR averaged over a large body region or the entire body, to the local 10g-averaged SAR. Both SAR quantities are limited according to IEC guidelines; in practice, the local SAR becomes more important at UHF because it typically reaches the limits first.
Although the total RF power fed to the RF coil can be measured during MR imaging and thus can provide an estimate for the global SAR, the local SAR cannot be assessed in vivo with sufficient accuracy. Therefore, electromagnetic simulations are conducted for different human body models and the spatial distribution of E and B1+ field are calculated as well as resulting SAR values are derived.
Further challenges at high and ultra-high field are related to the increase susceptibility that creates strong ΔB0 particularly at air-tissue boundaries. At 7T, ΔB0 values corresponding to frequencies of >400 Hz are observed in the brain above the nasal cavities and within the temporal lobes. This typically leads to signal reduction or signal voids in gradient echo images. Delta B0 also impacts balanced steady-state free precession (SSFP) imaging, which shows banding artifacts in regions with strong ΔB0. This effect is frequently observed in body imaging at 3 Tesla (18). Furthermore, B0 inhomogeneity also creates distortions, particularly in single shot acquisitions such as EPI. To limit this effect, acquisition time needs to be reduced by parallel imaging or other methods.
Another substantial practical challenge at UHF is the ability to detect the cardiac cycle using an electrocardiogram (EKG). The magneto-hydrodynamic (MHD) effect induces an elevated T-wave in the EKG signal, which in practice often leads to false detection of the R-peak (19).
Variations of the FA can be reduced by using adiabatic RF pulses; their resulting FA is substantially less prone to variations of the B1+ amplitude compared to standard RF pulses. However, a limitation of these pulses is the fact that they typically require higher RF power. Another promising solution for this problem consists in parallel transmission of different RF pulses on multiple transmit (TX) coil elements (20, 21). In B1+-shimming (3) all coil elements share the same RF pulse shape, but individual (time-constant) transmit phases and/or amplitudes are set for each RF coil element in such a way that the spatial B1+ variations are minimized. This approach has proven to be successful in a large range of applications and targets, particularly for targets within the body (22). In ‘true’ parallel transmission (pTX) or ‘transmit SENSE’ N independent RF pulses are applied to the N TX coil elements. This approach is particularly beneficial for localized excitations. An intermediate solution consists of ‘spokes pulses’ for the slice selective case and ‘kt-pulses’ for the non-selective case (23, 24). Both pulses consist of composed, B1+-shimmed sub-pulses (e.g. SINC or RECT pulses) together with gradient blips in between the sub-pulses. The combination of blips and B1+-shimming allows for improved homogeneity within the target volume compared to B1+-shimming alone.
RF power and SAR can be addressed in several ways. Stretching a given RF pulse by a factor f reduces the RF power by 1/f. Similarly, the VERSE principle (25) can be applied, which allows stretching only those parts of the RF pulse with high RF amplitude. State-of-the-art pTX RF pulse designs include the global SAR, the local SAR as well as the RF peak power as a constraint (26). In practice, a tradeoff between SAR/power and excitation fidelity needs to be found.
ΔB0 is typically reduced by B0 shimming; typically second-order shimming is performed but some systems are equipped with shims of even higher orders. Residual B0 variations can be addressed for example by shortening TE or by segmenting readout trains. On the excitation side, ΔB0 can be incorporated into the RF pulse design if necessary.
Increased imaging speed that is needed to address longer acquisition times due to higher spatial resolution is addressed using parallel imaging, by kt-acceleration (e.g. kt-blast or kt-GRAPPA), by compressed sensing or by applying Multiband (6-8).
Body applications at ultra-high field are still challenging, in this case a single universal shim setting (such as the CP mode for the head) that can be applied for all body targets and body shapes, does not seem to exist. However, the so-called ‘TIAMO’ technique combines two complementary shim solutions, which is a practical and robust approach (27). Body applications that require cardiac triggering are more difficult to perform at 3T and particularly at UHF because of the MHD effect. A solution to this problem was proposed by Frauenrath et al (28), who presented an acoustic cardiac triggering system that operates similar to a stethoscope.
Applications
While clinical 3T systems were already applied for head imaging soon after their introduction, the transition from 1.5T towards 3T for body applications such as the heart has been rather slow. Although the time scale is larger, a similar trend can be observed for applications at 7T. Currently most studies are targeting the brain while body applications remain rare.
Most applications aim for higher resolution or stronger contrast using structural, functional or spectroscopic imaging. Applied techniques which appear to benefit most from UHF are angiographic techniques, such as TOF, susceptibility weighted imaging and QSM, gradient and spin echo functional MRI, diffusion weighted imaging, spectroscopic MRI and X-nuclei imaging (2, 14-16).
These techniques have been applied to a large range of diseases and disorders, including tumor imaging, microvascular diseases, multiple sclerosis, epilepsy, dementia, Alzheimer’s disease, Parkinson’s disease and other neurodegenerative diseases (14, 29-32). Functional MRI at 7T in combination with high acceleration factors using parallel imaging and multiband allows for whole-brain fMRI imaging with 1mm isotropic resolution (33). Similar protocols are applied in the Human Connectome Project as well as in other functional studies performed worldwide.