Research participants:
B. van den Bergen,
C.A.T. van den Berg,
A.L.H.M.W. van Lier,
A. Raaijmakers and
J.J.W. Lagendijk In MRI antennas around a patient (figure 1) create a time-varying magnetic field in the patient. This field is also known as the RF (radio frequency) field and oscillates at the Larmor frequency of the precessing hydrogen nuclei (spins). These spins absorb some of the energy of this magnetic field and after some time the hydrogen atoms will emit this excess energy in the form of a signal. In this signal information over the spatial location and its environment (local anatomy) is entangled. After the signal is picked up by the antennas around the patient an image of the patient is reconstructed displaying the whole anatomy.
Figure 1, RF antennas (dots) around the patient inside an MRI magnet
As the magnetic field strength increases, the signal-to-noise improves. This can be used to increase spatial resolution and/or imaging speed (figure 2). Moreover, the high magnetic field strengths create new imaging contrasts which can display physiological information (figure 3).
Figure 2, difference in spatial resolution between 1.5 and 7 Tesla.
Figure 3, brain image with oxygen level overlay.
The increase in magnetic field strength poses however also some technical challenges. The excitation field becomes much more heterogeneous (figure 4), which gives image artefacts and the RF energy deposition in the patient increases considerably, which causes concern for patient safety (figure 5). Currently, we are working on solutions to overcome these problems.
Figure 4, homogeneity of the RF excitation field at 3 and 7 Tesla.
Figure 5, energy deposition in a patient at 3 and at 7 Tesla.
Phase-amplitude steering of the antennas to shape the radiofrequency field The time varying electromagnetic field in the patient that is used for signal creation is created by a so-called RF coil. This coil basically consists of a series of antenna elements, which traditionally all have fixed mutual phases and amplitudes (quadrature settings). Recent new coil designs have made it possible to alter these phase-amplitude relationships amongst the antennas for individual patients, enabling changes in the electromagnetic field towards more favourable field distributions. This changing of the phases and amplitudes is known as RF shimming.
The phase and amplitude settings will influence both the electric and magnetic field components. Settings that are favourable for the magnetic field might be unsuitable for the electric field and vice versa. This means that both can not be optimized individually, but rather that the optimization will need to improve the electric and magnetic parts of the field simultaneously. This makes the optimization process far from trivial, and typically a trade-off is found between the homogeneity of the magnetic field on one side and the amount of heating caused by the electric field on the other side (figure 6).
Figure 6, Effect of RF shimming (see text). Quadrature settings are traditionally used. The excitation field can be made homogeneous in the central ellipse, while the energy deposition is reduced. A trade-off is found between excitation field homogeneity and energy deposition.
Rapid online electromagnetic field computations We participated in the 'Study group mathematics with industry 2007' (www.math.uu.nl/swi2007) by providing a challenging case. Together with a group of enthusiastic mathematicians a new method was developed that can be used for fast electromagnetic field calculations.
Information about the electromagnetic fields in the patient can be used for RF shimming, where both the excitation field can be shaped and the energy deposition minimized. However, traditional simulation methods take several hours to provide this type of information, which limits their use to research purposes. The new developed method provides the same information in a single minute (figure 7), thereby opening the door for RF shimming to be implemented in a clinical setting.
Figure 7, electric and magnetic field calculation with the old (a, c) simulation method that takes 12 hours, and the new method (b, d) that takes 1 minute.