Nuclei with an odd mass or odd atomic number have an intrinsic spin angular moment with an associated magnetic moment. In the presence of an external magnetic field B₀, the magnetic moment of nuclei with a one half spin will line up in parallel with the applied magnetic field with an either spin-aligned or spin-opposed orientation. The energy difference between the two orientations is very small at room temperature with a small majority of nuclei populating the lower energy spin-aligned orientation. This will cause a net macroscopic magnetization M₀ which will be directed parallel to B₀ and which strength will depend on the spatial spin density and the magnetic field strength. Next to this magnetization, nuclei will tend to precess in phase around the magnetic field with a frequency depending on the nucleus, the molecular structure, and the strength of the magnetic field. This frequency is traditionally called the Larmor frequency. Generating a MR signal is a two-phase process. During the transmission phase, a radiofrequency (RF) signal induces a “spin flip” resulting in a rotating transversal magnetization in the plane perpendicular to the longitudinal external magnetic field B₀ and in a residual longitudinal component if the flip angle (FA) is not 90°. During the receiving phase, the transversal magnetization is picked up to generate the MR signal. During transmission, the RF excitation frequency must fit the Larmor frequency to have optimal resonance conditions, hence the term magnetic resonance. The Larmor frequency for the ¹H nucleus amounts to 42.58 MHz/Tesla.

After excitation and during the receiving phase, relaxation occurs when magnetization returns to the state prior to RF excitation. One can differentiate two types of relaxation. Longitudinal relaxation describes the relaxation where the longitudinal magnetization component is exponentially restored to initial net longitudinal magnetization M₀. This relaxation is referred to as spin-lattice relaxation and is quantified by the longitudinal relaxation time constant T1. Transverse relaxation corresponds to the so-called spin-spin relaxation and describes the exponential decrease of the transverse magnetization component perpendicular to the longitudinal external magnetic field. The representative parameter is the transverse relaxation time constant T2. Both relaxation processes take place simultaneously with relaxation times T1 and T2 being characteristic for a specific tissue type. While the T1 relaxation time is within the range of 300 ms (fat) to 3,000 ms (fluids), T2-relaxation times are always much shorter ranging from 30 ms (muscle) to 2,000 ms (fluids) (Bazelaire et al. 2004). Specifically for brain imaging and for a 3 T magnetic field, T1 relaxation times are 832 msec for white matter and 1,331 ms for gray matter, while T2-relaxation times are 79.6 and 110 ms, respectively (Wansapura et al. 1999).

MRI consists of applying the NMR principle to construct an image representing the proton density of different tissue voxels. Many of those protons are water protons such that MRI is particularly well suited for imaging soft tissue structures. The frequency of the proton MR signal is proportional to the magnetic field to which they are subjected during relaxation. An image can be made by producing a set of well-calibrated magnetic field gradients in all directions so that a certain magnetic field strength can be associated with a given location. In the longitudinal direction, a slice selective excitation is induced by applying a magnetic field gradient, a so-called slice selection gradient. This way, Larmor frequency and RF excitation fulfill the resonance condition in a specific axial slice such that only the magnetization of that slice is excited and contributes to the signal. In one of the transversal directions, phase encoding is applied with different phase-encoding gradients between RF excitation and readout. This is achieved by briefly switching on a magnetic field gradient in that direction when neither excitation nor readout is performed. This additional magnetic field gradient will induce, differences in spinning rate such that when the gradient is switched back off again, proton spinning is dephased along this direction and the signal is phase encoded. Finally, frequency encoding is applied along the other transversal direction by applying a readout gradient. This way, the frequency of the signal during readout is dependent on the spatial location in that direction. The signal of all voxels from the selected slice is recorded simultaneously in the so-called k-space by performing sequentially different phase-encoding steps. This k-space represents the Fourier transform of the MR image. The central part of k-space covers the lower spatial frequency range of the image representing the contrast and signal-to-noise content, while the outer part of k-space contains the high frequencies or resolution content. In case of contrast-enhanced MRI, the central part of k-space could be sampled first using a more centric approach, while for very fast scan techniques, spiral-like trajectories could be considered instead of the line at a time strategy. Multi-slice imaging and exploiting k-space symmetry can further reduce the acquisition time substantially.

During the phase encoding, phase is also encoded outside of the field of view (FOV), such that structures or body parts that are positioned outside of the FOV but within the encoded image slice contribute to the image and can cause aliasing artifacts in the phase-encoding direction. Other artifacts that mainly propagate in the phase-encoding direction are motion artifacts due to the movement of spins between the phase encoding and signal reading. During movement through the magnetic gradient field, spins change frequency and thus acquire an additional phase shift causing a spatial mismapping of the signal of these protons. In the frequency-encoding direction, the pace of spatial encoding and sampling is such that physiological motion only causes limited spatial blurring in that direction.

In contrast, chemical shift artifacts can arise in the frequency-encoding direction. The chemical bounding of protons causes a shift in their precession frequency due to magnetic shielding provided by the electron shell. For the same field strength, protons bound to water have higher resonance frequencies than protons bound to lipids. Therefore, spatial encoding by the frequency-encoding gradient will shift fat relative to water in the frequency-encoding direction resulting in dark and/or white band at the interfaces between water and fat. This chemical shift artifact can be reduced by suppressing the signal from fat using saturation or inversion recovery techniques. Another method to minimize chemical shift artifacts is to use a wider receiver bandwidth, such that the bandwidth for each pixel is wider and thus a given chemical shift corresponds to fewer pixels. Drawback of wider receiver bandwidth is however a lower signal-to-noise ratio.

Following the excitation pulse, the net magnetization in the transverse plane will start to dephase due to T2 relaxation and signal intensity will drop very fast. A short time after the excitation pulse, a 180° pulse is transmitted causing the spins to rephase and the signal to rise again. This signal is sampled and called an echo of the initial net transversal magnetization. Echo time is twice the time between the initial excitation RF pulse and 180° RF pulse. The initial excitation pulse used to induce a flip angle (FA) of 90°; however, smaller or larger flip angles are used as well. Advantage of a SE sequence is the strong signal which is compensated for local field inhomogeneities and different chemical shifts and is therefore less susceptible to artifacts. Rephasing however requires longer minimal echo times due to the additional 180° pulse and longer repetition times for an acceptable signal. This will increase total scan time and therefore the risk for motion while a larger amount of RF energy will be deposited in the body. In order to reduce acquisition time, multiple echoes are generated after a single excitation pulse using an echo train (Harms et al. 1986). The sequences are called turbo spin echo (TSE) or fast spin echo (FSE) sequences. The number of echoes is called the turbo factor, and measurement time is reduced by this factor compared to standard SE imaging. However, contrast in fast spin echo needs to be described differently compared to standard spin echo sequences since multiple echoes are generated at different echo times. From these echoes, the ones corresponding to the central part of the k-space will determine image contrast, and the time between these echoes and the excitation pulse is called the effective TE.

Since a 180° rephasing pulse is used for the echo, SE and FSE sequences are insensitive to microscopic magnetic field inhomogeneities caused by iron deposition or specific tissue composition. Therefore, real T2 weighting is only possible by this kind of sequences, excluding T2* contrast and opposed-phase imaging. On the other hand, SE and FSE sequences generate a strong signal compensated for local field inhomogeneities and therefore less sensitive to artifacts. Examples of T2 weighting can be found in Figs. 5.2 and 5.3.

GE sequences differ from SE sequences in the way the echo is formed. GE sequences create an echo by reversing the polarity of the readout or frequency-encoding gradient. A gradient pulse directly after the RF pulse enhances dephasing of spin frequencies such that the FID occurs considerably faster than under normal conditions. By reversing the polarity of the gradient, dephasing of the spins is reversed and spins start rephasing till at a certain moments all spins are in phase again and the echo is generated and measured. This gradient polarity reversal technique can work much faster than a 180° pulse for rephasing spins. This technique can be combined with a partial FA, which means smaller than 90°, making this approach very suitable for fast imaging sequences. Since a partial FA decreases the amount of magnetization flipped into the transverse plane, longitudinal magnetization is recovered faster, allowing a shorter TE and TR and thus decreased scan times while preserving image quality and limiting RF deposits. On the other hand, GE sequences are susceptible to microscopic magnetic field inhomogeneities and chemical shift effects. GE imaging sequences therefore produce T2*-weighted images reflecting faster spin dephasing and signal loss due to both spin-spin interactions and magnetic field inhomogeneities and allowing opposed-phase imaging. Due to different resonance frequencies of water and fat, the signal intensity from a voxel containing both water and fat will oscillate, reaching a minimum when fat and water are out of phase and a maximum when they are in phase. If the echo is created with fat and water in phase, their signals will add up. When the echo is performed when they are out of phase, their signals will be subtracted. This way, water and fat can be separated. For spin echo sequences, these effects are not feasible since phase shifts due to chemical shift are canceled by the 180° refocusing pulse.

EPI sequences are a very fast way of acquiring MRI data and form the basis for many advanced MRI applications such as diffusion, perfusion, and functional imaging. The excitation pulse is followed by a gradient echo train to continuously measure the MR signal and fill the k-space completely (single shot) or partially (segmented acquisition) within a single TR interval. The gradient echo train consists of fast oscillating gradients for frequency encoding in the readout direction and short gradient pulses for phase encoding. Therefore, a high-performance gradient system is required with short rise times to allow frequent gradient switching. Specific compositions of readout and phase-encoding gradients define the spatial encoding and the trajectory to fill up k-space. Several types of trajectories such as spiral trajectories can be compiled to optimize sampling of k-space and reduce scanning time. However, this often results in poorer spatial resolution. Spatial encoding of EPI sequences is sensitive to magnetic susceptibility and can be disturbed by gradient imperfections. Increasing the scan time and using segmented rather than single-shot sequences can reduce artifacts. Due to the narrow readout bandwidth in the phase-encoding direction, chemical shift artifacts are possible in this direction requiring fat signal suppression.

The contrast generated with an EPI sequence is determined by the echo sequence and additional contrast-related gradient pulses. Both SE-EPI and GE-EPI are available.

In case of SE sequences, the effect of T2 relaxation on the image is controlled by TE, while the impact of T1 relaxation is dependent on TR. For a sequence with a short TR and very short TE, spins will undergo very limited dephasing due to T2 relaxation, and without T1 relaxation all tissue types would generate a similar MR signal. However, the short TR will not allow the net magnetization in the longitudinal direction to recover fully between two RF excitation pulses. The recovered longitudinal magnetization will depend on the T1 relaxation time of the underlying tissue type. Fat for instance with a shorter T1 relaxation time will have a larger net longitudinal magnetization restored than water with a longer T1 relaxation time. Consequently, the transversal magnetization generated during the next RF excitation pulse will be higher for fat than for water, and therefore, the contribution of fluids to the overall signal will be lower. In this case the MR signal is T1 weighted, meaning that contrast is dependent on T1 relaxation differences between tissues where tissues with longer T1 relaxation times contribute less to the signal and therefore appear darker in the MR image.

If the TE of a SE sequence is prolonged, dephasing of spins due to T2 relaxation will be more pronounced and will depend on the differences in T2-relaxation times between tissue types. If a longer TE is combined with a relatively long TR, full recovery of the net longitudinal magnetization is obtained for all tissues and tissue differences in T1 relaxation are canceled out. Consequently, the transversal magnetization generated during the next RF excitation pulse is the same for all tissues such that the MR signal will be dependent only on differences in T2 relaxation and the MR image contrast will be T2 weighted. Water for instance has a long T2-relaxation time and will still show some phase coherence in case of a long TE, while most other tissues will have dephased much more and will contribute less to the MR signal. Hence, water will appear bright in a T2-weighted image, while other tissue types will have lower intensities (see Figs. 5.2 and 5.3).

If a sequence combines a short TE with a long TR, T2 relaxation will have a very limited impact on the image contrast, while the net longitudinal magnetization will have recovered for most tissue types. Thus, the MR contrast will be dominated neither by T1 relaxation nor by T2 relaxation but will depend only on the proton density (PD) of the different underlying tissue types. A tissue type with a high proton density such as water will produce a high MR signal and appear bright, while tissues containing few protons will generate a low signal and appear dark.

SE sequences are especially suited for acquiring images of the same slice with different contrast. Generating a second echo with a longer TE within the same TR interval and with the same phase encoding will generate a PD-weighted image for the echo with the short TE and a more T2-weighted image for the echo with longer TE. This way, a T2- and PD-weighted image of the same slice position can be acquired without increasing overall acquisition time. Specifically for FSE sequences, a short TR for T1-weighted sequences limits the echo train length. Therefore, FSE sequences are preferably used for T2-weighted imaging.

Contrary to SE sequences where the FA is mostly fixed to 90°, GE sequences use a range of flip angles and contrast is determined mainly by FA and TE. If FA decreases, T2* weighting is favored since the residual longitudinal magnetization will be higher and the recovery of longitudinal magnetization for a given T1 and TR will be more pronounced. A high FA and short TE will generate T1 weighting, while a medium FA and short TE corresponds to PD-weighted contrast.

Given these types of contrast, it is not easy to decide on the best image contrast for a given pathology. In general, T1-weighted contrast allows a clear delineation of anatomical brain structures, while for brain pathologies involving edema, T2- or PD-weighted contrast could be preferred (Fullerton 1984).

Inversion recovery (IR) sequences are frequently used to generate T1-contrast-weighted images. IR sequences can be considered a standard sequence preceded by a 180° excitation pulse in the longitudinal direction. This 180° excitation pulse will induce T1 relaxation that takes twice as long as for a standard sequence. No T2 relaxation will occur since no transversal magnetization component is generated. After a specific time interval of T1 relaxation, called the inversion time TI, a standard sequence is applied to generate an image. Because of this preceding 180° excitation pulse, one can make better use of differences in T1 relaxation between various tissues to create higher T1-weighted image contrast. This is however at the expense of longer acquisition times compared to standard T1-weighted sequences. TR time is now defined as the time interval between two 180° excitation pulses in the longitudinal direction. Since TE is relatively short compared to TI, image contrast of IR sequences is mainly determined by TI. Choosing this TI appropriately will suppress the signal from specific tissue types. If for instance the standard sequence part of an IR sequence is started at the time the longitudinal magnetization vector of fat or cerebrospinal fluid (CSF) crosses the zero line, no magnetization can be flipped into the transversal plane and therefore no signal can be generated for this tissue type. In case of fat, this type of IR sequence is called short TI inversion recovery (STIR) (de Kerviler et al. 1998) and is very effective in those cases where fat signal interferes with the given pathology such as the detection of water in fatty tissue in case of inflammation in the joints or infiltrating tumors in bone marrow. The IR sequence for suppressing the signal of CSF is called fluid-attenuated inversion recovery (Eastwood and Hudgins 1998; Simonson et al. 1996) (FLAIR) (see Fig. 5.2). This sequence is very useful for visualizing demyelinating diseases and to assess vascular white matter hyperintensities and infarctions.

The uniformity of the static magnetic field B₀ is slightly distorted by the magnetic susceptibility or magnetizability of substances that are present in the FOV. This magnetic susceptibility exhibited by a tissue or substance will result in signal loss (Wendt et al. 1988). In particular the magnetic susceptibility of oxygenated and deoxygenated blood differ, with oxygenated blood being diamagnetic and decreasing local field strength, while deoxygenated blood is paramagnetic and increasing local field strength. Protons in the tissue surrounding the blood vessels are affected by this microscopic field variation, resulting in a loss of phase coherence. This induces signal cancelation and a decrease in total signal intensity. Next to phase changes at blood/tissue interfaces, the difference in magnetic susceptibility between oxygenated and deoxygenated blood, although small, also induces T2* changes between the arterial and venous bed of the vascular system. Using 3D GE sequences, susceptibility-weighted imaging (SWI) picks up these subtle differences in magnetic susceptibility in order to detect bleedings and small iron depositions (Chavhan et al. 2009) (see Fig. 5.3) or as part of a multimodal imaging approach for neurodegenerative diseases (Haller et al. 2010) such as Alzheimer’s (Ramani et al. 2006) and multiple sclerosis (MS) (Langkammer et al. 2013; Tavazzi et al. 2007; Walsh et al. 2014).

Diffusion reflects the Brownian, random movement of water molecules at microlevel. This random motion is hindered by cell membranes, fibers, or large molecules such as proteins, while concentration of these obstacles depends on tissue type or pathology (tumor tissue, abscess, and edema). Measuring the mobility of water molecules therefore gives indirect information about the tissue structures in terms of cell density and granularity such that tissue types can be better characterized and differentiated.

The contrast of diffusion-weighted imaging (DWI) is based on the mobility of the extracellular water by adding diffusion gradients to the imaging sequence. The spins of immobile water molecules will dephase by the first gradient pulse but will rephase by the second gradient pulse such that no impact on the signal is observed. Contrary, the spins of water molecules moving along the direction of the diffusion gradients during the time interval between the gradients will not be rephased by the second gradient and will stay dephased compared to the spins of the immobile water molecules and thus the signal will decrease. The more water molecules diffuse during the time interval between the two diffusion gradients, the more they will remain dephased after the second gradient and the more the recorded signal will be weakened. Weakening of the signal is limited to diffusion along the direction of the gradients, so diffusion gradients are applied in three perpendicular spatial directions. The diffusion magnitudes calculated along these three directions are averaged to provide a measure for a global diffusion-weighted image called a trace image.

DWI sequences are usually T2-weighted ultrafast 3D SE-EPI sequences sensitized for diffusion by adding diffusion gradients symmetrically before and after the 180° rephasing pulse. Therefore, contrast of DWI images is both diffusion and T2 weighted such that a strong signal could correspond to diffusion restriction as well as T2 shine through of a T2-sensitive lesion. The diffusion gradient pulses are characterized by the b value which is essentially dependent on the gradient amplitude, gradient duration, and time interval between the two gradients. In clinical practice a T2-weighted image with a b value of zero is combined with a diffusion-weighted image with a high b value (see Fig. 5.4).

If DWI images are acquired for different b values, the amount and speed of movement can be quantified by the apparent diffusion coefficient (ADC) (Vandecaveye et al. 2008; Thoeny et al. 2012) (see Fig. 5.4). The ADC is calculated by fitting an exponential diffusion model to the signal intensities from images acquired with different b values. This way, only information on diffusion is retained and T2 shine through of T2-sensitive lesions is avoided. Whereas fast-moving molecules will quickly lose all of their phase coherence and signal intensity, even at low b values, slow-moving molecules will retain high signal intensities far into the higher ranges of b values. A solid tumor will show high signal intensity on a high b value image and low signal intensity on the corresponding ADC map, whereas apoptosis or necrosis will commonly appear as areas of low signal intensity on high b value images with high signal intensity on the corresponding ADC map.

This way, post-therapeutic changes induced by radiation therapy or chemotherapy generate a high signal intensity (edema) or low signal intensity (necrosis) on high b value images but generally show high signal intensity on corresponding ADC map. ADC information proves clinically relevant for other brain pathologies (Schaefer et al. 2000) such as degenerative (Vitali et al. 2011; Kim et al. 2011; Horsfield and Jones 2002) (Creutzfeldt-Jakob’s disease) or inflammatory pathologies (MS) (Tavazzi et al. 2007).

Generally, two approaches can be applied for perfusion-weighted MRI. The first approach uses endogenously generated contrast, while the second approach relies on the intravenous administration of MR-sensitive contrast agents. PWI information can be used to quantify brain blood volume and extract temporal information about brain blood flow on the level of microcirculation.

MRI sensitivity to movement in terms of spatial encoding perturbations and artifacts can be used to develop vascular imaging approaches based on signal changes that are linked to flow. Techniques such as time-of-flight MRA (TOF MRA) or phase-contrast MRA (PC MRA) apply endogenous vascular contrast, while contrast-enhanced MRA exploits the relaxation properties of exogenous contrast agents to visualize vascular structures (van Laar et al. 2006; Ozsarlak et al. 2004).

MR angiography techniques are generally T1-weighted echo-gradient-based sequences. When endogenous vascular contrast is used, strategies are implemented to suppress the background signal represented by the stationary tissue such that vascular signal is favored over that of the surrounding tissues (Foo et al. 2005).

TOF MRA generates contrast between blood flow and stationary tissue by changing the magnitude of magnetization such that moving spins generated a high signal while the signal of stationary tissue spins will be diminished. Using repetition times much shorter than the relaxation times of brain tissue, spins in a predefined slice do not fully recover after each repetition and the MR signal decreases with the number of excitation pulses until equilibrium is reached. Moving spins that flow into this slice are not saturated and therefore have a greater signal, creating contrast between saturated stationary tissue and blood. Drawbacks are the long acquisition time and saturation of in-plane blood vessels, while the signal of stationary tissues with short T1 relaxation time such as fat, hematoma, or thrombus can be poorly suppressed.

PC MRA on the other hand generates contrast between blood flow and stationary tissue by changing the phase of magnetization such that the phase of moving spins is shifted relative to the phase of stationary spins. This technique uses a bipolar flow encoding gradient of given intensity to dephase spins in the transversal direction in proportion to their velocity in the direction of the gradient. This generated phase difference between stationary and moving spins can be picked up to derive contrast between flowing blood and brain tissue.

Next to TOF MRA and PC MRA, time-resolved ASL (Robson et al. 2010) has been developed as a noninvasive alternative to angiography.

In the case of CE MRA (Zhang et al. 2007a), signal differences are achieved not only by employing the appropriate sequence parameters but also by intravenously injecting a contrast agent into the vascular system to selectively shorten the T1 relaxation of the blood. Using a T1-weighted imaging sequence during the first pass of the contrast agent, the shortened T1 relaxation of blood causes the blood to give rise to a very large signal producing a very high contrast between arteries and surrounding tissues and veins.

While MRA using endogenous contrast can be disturbed by complex or turbulent flows, the blood signal of CE MRA is minimally affected by dephasing caused by complex flows and susceptibility variations. The key factor for CE MRA signal is to acquire data at the right moment, during the passage of the contrast agent bolus. Because the CE MRA techniques are relatively insensitive to this signal loss, they provide high-quality images with fewer artifacts than the non-CE methods. Because effects of saturation are minimal, large fields of view can be imaged with high SNR to demonstrate large vascular areas in a short acquisition time. The short imaging time permits acquisition in a single breath-hold interval, providing high-quality images even in areas affected by respiratory motion.

Next to the MRI techniques which were mentioned above and which are used extensively in a clinical setting, other MRI techniques will be discussed in brief. These methods are mainly used for addressing scientific questions rather than for clinical diagnosis or therapy evaluation. However, for specific brain pathologies, they could move from a research setting into clinical tools in the near future.

Increasing the magnetic field strength yields images with a better signal-to-noise ratio (SNR), a higher spatial resolution, shorter acquisition times, and a better contrast for tissue characterization (Schick 2005; Stafford and Challenges 2003; Panebianco et al. 2013; Mri and Quick 2011). In terms of relaxation times, T1 relaxation time increases, while T2-relaxation time decreases. Therefore, MRI sequences require appropriate adjustment for the slower longitudinal relaxation, while a shorter TE is needed to comply with the reduced T2-relaxation time.

Both TOF MRA (Nguyen and Yang n.d.) and ASL benefit from the intrinsically higher SNR and the lengthening of blood T1 relaxation providing better contrast between tissue and the circulating blood. Increased contrast between tissue and paramagnetic contrast agents will allow a reduced injection dose of these agents. Higher sensitivity to magnetic susceptibility has a positive impact on the detection of hemorrhages, first-pass perfusion imaging, and BOLD fMRI signal. In terms of chemical shift, the larger gap between the resonance frequencies of fat and water makes saturation of the fat signal much easier and more homogeneous.

Besides these advantages, high-field MRI has some limitations as well (Schick 2005; Stafford and Challenges 2003). Due to the shortened wavelength of the RF pulse, now being much shorter than the FOV, RF excitation occurs much more inhomogeneous. On the other hand, the increased magnetic susceptibility can be diminished by reducing the TE and making the sequence less T2* sensitive or by choosing segmented rather than single-shot sequences and increasing the bandwidth. Finally, during high-field MRI, the amount of RF energy deposited in tissue is increased significantly and needs to be addressed by increasing TR, limiting the number of slices, reducing the flip angle, or shortening the length of echo train.

In this context, it is also worthwhile to briefly mention parallel imaging (Glockner et al. 2005; Deshmane et al. 2012; Setsompop et al. 2008). Multiple small diameter coil elements can be combined in a phased array to record signals simultaneously and independently. Each coil element has a limited receiving range and a sensitivity profile that depends on the distance from the coil element. Therefore, the signal received by each coil element contains spatial data such as position coil position, reception level, and sensitivity level. This information can be combined with the gradient-induced spatial encoding to reconstruct the image. Since small diameter coils have higher signal-to-noise ratios than coils with larger diameters and noise between the different elements in the phased array is not correlated, the signal will have a better SNR than that one large coil.

Techniques combining the signals of several coil elements in a phased array to reconstruct an image are called parallel acquisition techniques. These techniques can be used to increase SNR, improve image quality for the same acquisition time, or reduce the acquisition time. The latter is particularly suitable for breath-hold sequences or for improving the temporal resolution of perfusion imaging (Oliver-taylor 2012), functional imaging, and dynamic imaging of movements while reducing flow and motion artifacts. Especially cardiac and abdominal MRI or MRA (Wilson et al. 2004) can benefit from this approach.

In terms of reconstruction, two approaches have been presented (Blaimer et al. 2004). The first approach combines the images from each coil to reconstruct the global image in the spatial domain (f.e. SENSE or sensitivity encoding) (Zhang et al. 2007b) and is the most widely spread in present commercially available parallel MR sequences, while the second approach combines the frequency signals of each coil in the Fourier domain (GRAPPA or generalized auto-calibrating partially parallel acquisition) (Hoge and Brooks 2008).

In the context of ultrahigh-field MRI (Wiesinger et al. 2006), parallel transmission techniques, where the independent RF pulses from an array of small coils can be combined to a more complex and adaptive RF pulse, offer multiple potential improvements in radiofrequency transmission (Katscher et al. 2003; Katscher and Börnert 2006). More advanced RF transmission can lead to shorter pulse duration and reduce energy disposition in the patient, while patient-specific inhomogeneities can be corrected for, especially in ultrahigh-field MRI. Therefore, advanced parallel acquisition techniques allow sequences of the echo-planar type and ultrafast gradient echo type, even at a very high field (Webb and Collins 2010; Harvey and Hoogeveen n.d.).