Basic Principles of Magnetic Resonance Imaging (MRI) and Spectroscopy (MRS)One

The first experiments of nuclear magnetic resonance (NMR) were performed in 1946 and since then its application in physics, chemistry, and biology has been phenomenal. Magnetic resonance (MR) is a technique for probing atoms and molecules based upon their interaction with an external magnetic field (B₀). Initially NMR was used for chemical structure elucidation, and its potential related to medicine was recognized almost at the same time. The most familiar application of NMR phenomenon to the common man and the clinical community is magnetic resonance imaging (MRI), which is used to produce anatomic images for diagnostic purposes of various pathologies. In 1973, high-resolution phosphorus (³¹P) NMR studies of intact blood cells¹ and later on muscle tissues were reported^{2, 3}. Ackerman et al⁴ reported ³¹P NMR study of skeletal muscle and brain metabolism noninvasively in small animals using a surface coil. With the availability of larger magnets extension to human ³¹P MR studies, followed later.

NMR is the magnetic interaction between nuclei of atoms and radio-frequency (RF) field (B₁) in the presence of an external magnetic field (B₀). According to quantum mechanics, a subatomic particle such as proton has quantized angular momentum, called spin (I). Associated with spin is a magnetic moment. Hydrogen atom has only one proton as nucleus and hence this property can be observed by looking at hydrogen. The human body consists mostly of water and fat and since hydrogen is a constituent atom in water and fat, we have a potential medical application. Since each water molecule has two associated hydrogen atoms covalently bound to an oxygen atom, there are approximately 5 × 10²⁷ hydrogen nuclei in an adult human body.

These spins (magnetic moments), align in an externally applied field (Fig.1.1) which can be treated as a macroscopic magnetization (M₀) following the laws of classical mechanics. This macroscopic magnetization align in the direction of the applied field B₀, usually referred to as the z–direction and denoted as M_z (=M₀). More spins will line up in parallel orientation, since it is the lower energy state. The energy difference between these two levels is given by

where, ν₀ = (γ/2π)B₀, is the frequency of electromagnetic field. Applying a magnetic field B₁ perpendicular to the main static field cause rotation of the macroscopic magnetization. That is, with the application of a RF field, the nuclei are flipped from a parallel to an anti-parallel state. After the pulse, the nuclei return to the parallel state emitting RF energy, which is detected. This relationship is readily expressed by a simple Larmor equation

where ω₀ (=2πν₀) is the Larmor precessional frequency of nuclei in megahertz (MHz), γ the gyromagnetic ratio which is different for different nuclei and B₀ the external magnetic field in Tesla (1 Tesla = 10⁴ Gauss).

The magnetization arising from different number of nuclei (say proton) present in tissues is the so-called net (or bulk) magnetization M₀ (=M_z) which is parallel to the external magnetic field, B₀. A strong radio-frequency field B₁ applied along x-axis, tips M₀ away form z-axis (Fig. 1.2A). The duration and power of the RF pulse determines the direction of M₀ after the pulse. If a so-called 90°_x or π /2 pulse is applied, M₀ points along the positive y-axis (Fig. 1.2B), i.e. the z-magnetization (longitudinal) is now transferred into a transverse plane. The Larmor frequencies of the various (proton) nuclear magnetic moments vary and as a consequence, the vector M (actually M_xy) now splits into its components (Fig. 1.2C) in the xy plane and produce a voltage signal. A plot of the time dependence of the intensity of the y-component of the M_xy magnetization (that is, the voltage induced in the receiver coil) as a function of time is known as free induction decay, FID (Fig. 1.2G). Thus, FID represents the time evolution of the transverse magnetization. It is necessary to analyze this time-domain signal f(t), into its individual frequency components, F(ω). This is accomplished by a mathematical formulation, known as the Fourier transformation (FT), which is given by

With the RF switched off, the macroscopic magnetization will get back into the original position by a process called relaxation. Initially, the nuclei are in thermal equilibrium with the static field B₀ (along z-direction magnetization, M₀ = M_z) when they are completely aligned (Fig. 1.2A). After RF excitation, the nuclei are disturbed and have tendency to get back to the original thermal equilibrium by realigning (Fig. 1.2F) with B₀. The time constant that describes the rate, at which the z-component of the net magnetization (M_z) returns to its equilibrium value M₀, is the T1 relaxation time. This happens due to the excited nuclei losing its energy to the surrounding molecular environment, called the ‘lattice’, hence spin-lattice relaxation. The recovery of z-component of the magnetization (M_z) is given by6

where M₀ is the initial magnetization and ‘t’ is the elapsed time from the start of the FID. In MRI, signals are repeatedly generated by the application of a series of 90° RF pulses, which establish a ‘steady state’ magnetization. With the constant repetition time (TR), the signal is given by

This effect is called ‘saturation’ and is the basis of T1 contrast.

MR images are produced by assigning shades of gray, white, and black to the strength of the signal produced by the relaxing protons. Diffusion-weighted imaging (DWI) is a technique that generates tissue contrasts that are based on completely different phenomenon than those that are involved in conventional MRI technique. Diffusion techniques are based on the amount and direction of Brownian movement of free water protons; further details are presented later in this article.

The quality of MRI systems has been available since the early 1980s have improved exponentially in recent years. Although the early systems used resistive magnets, super conducting magnets are now standard. MRI scanners are generally characterized by the strength of the magnetic field. Most imaging procedures are performed with field strengths in the range of 0.2 to 1.5 Tesla, although imaging outside this range is possible. Larger field strengths permit shorter imaging time and improved resolution. The strength of the magnetic field determines the tissue (proton) resonant frequency. This is the frequency that is receptive to the RF pulses applied to the tissue and is also the frequency of the RF signals emitted during the imaging process.

where ω_i is the frequency of the proton at position r_i and G is the vector representing the gradient amplitude and direction. It is, therefore, easy to see how a 1-D image projection is made. The signal is acquired in the presence of a field gradient and the intensity is plotted against the frequency⁶, with the frequency axis being ‘coded’ for, i.e. be proportionately equivalent to the spatial position. Images in two or three dimensions are made by an extension of this coding concept, using additional magnetic field gradients along the y- and z-axes, and recognizing that a precessing spin experiencing a gradient pulse of duration ∆t along either of these axes will accumulate an additional phase, e.g.,

A pulse sequence is a series of timed events involving RF pulses, gradient pulses and digital sampling of an analog signal. It accomplishes two tasks: (i) to collect the data in an orderly fashion so that the origin of the signals can be determined – i.e., the pixel 10position – and this is the function of the gradient magnetic fields outlined earlier; and (ii) it influences the image contrast – pixel character – by specifying the timing and power of the RF pulses. Many of the parameters that have to be specified in a pulse sequence, such as the timing and magnitude of gradient magnetic fields are included in the computer software.

where C is a constant related to the amplification of the MRI receiver, and M_z reflects the proton number density. For short echo times (TE), eqn. (9) simplifies to

Multislice and multiecho techniques help to decrease total scan time.

Measurement times are long due to incorporation of long TR values plus the inversion time (TI) component in to the sequence. Tissue signal suppression may be observed using STIR (short tau inversion recovery), a special case of IR sequence. By using a short TI, images with suppression of signals from specific tissue types, such as fat, can be achieved. In STIR, the idea is to wait only until the fat signal crosses the null point (0.69 × T1), the time at which there is no fat signal available to be flipped to the transverse plane. A variation of STIR is FLAIR (fluid attenuated inversion recovery) and in this sequence, rather than using a short TI value, a very long TI is used to catch water at the null point. This results in suppression of structures such as ventricles (CSF) and has been shown to help to define even very small demyelinating lesions such as multiple sclerosis (MS). The drawback of FLAIR lies in the fact that TR value greater than 5000 ms are required with TI values of about 2000 ms, thus increasing the measurement time.

The SE pulse sequence has been the fundamental platform for MR imaging. However, sometimes measurement times are long and the sequence is sensitive to motion artifacts and thus becomes unsuitable for many sick patients. There are several reasons to acquire MRI data quickly and the principal impetus is to improve clinical efficiency. In addition, to minimize patient discomfort and the adverse effects of involuntary physiological patient motion on image quality. Finally, shortening of scan time can open many new avenues for dynamic scanning such as cardiac imaging, functional studies like task activation, etc. The methods for reducing the measuring time as proposed by various groups can be divided into two categories. First, using gradient echoes and the second one based on a different phase coding of signals (echoes) recorded within one pulse sequence (mostly CPMG sequences) after one single excitation of the spin system. GE sequences use a small flip angle for excitation, followed by a gradient reversal instead of the 180° RF pulse for echo formation, which is applied in the read out direction (Fig. 1.6). These methods require highly homogeneous B₀ field. By employing gradient reversal it is possible to achieve shorter echo times of the order of TE = 1 or 2 ms. In addition, it is possible to use a short flip angle which helps to use shorter repetition times. FLASH (fat low angle shot)²² is one such sequence, which uses a flip angle of 15 to 30° instead of a 90° pulse. Another method that is also based on GE is FISP (fast imaging with steady state precession)²³. In FISP the slice gradient, the phase encoding gradient and the read gradient are always switched on a second time with equal duration but opposite sign, so that finally all spins are again in the same position they were at the beginning. In this way, dynamical steady state is reached for the longitudinal as well as transverse magnetization. Also, GE sequences have increased contrast sensitivity over SE sequence due to increased T2^* dependence.

EPI is a ‘snap shot’ sequence to map the multiple phase-echoed lines of k-space with a single RF excitation, i.e. a complete data set is acquired to reconstruct an image from the MR signal produced by a single RF excitation of the selected slice. The original concept and its early realtime biomedical application employed a continuous scan path, zigzagging rectilinearly through the entire k-space plane. This 13requires a clever manipulation of the encoding gradients Gf and Gp. Gf is inverted after each line-scan while Gp is applied very briefly in ‘flips’ after completion of a line. The inversion of a gradient actually inverts the scan direction in k-space. The gradient Gf is strong and rapidly switched to obtain very fast echoes, while the gradient Gp is flipped as narrow pulses coinciding with the zero crossings of Gf. These gradients are programmed to act in synchrony, generating a fast zigzagging route through k-space. The MR signal is continually sampled during this k-space traverse, and the method is therefore very fast and efficient.

Most of the fast imaging sequences in use like EPI, FLASH, etc achieve their speeds by optimizing the gradient strengths, the switching rates, and the patterns of applied gradients and pulses. Beyond a certain threshold, however, rapidly switched gradients are known to produce neuromuscular stimulation, while excessively dense RF pulse trains can lead to unacceptable levels of RF energy deposition and heating of the tissue. The common feature of all the fast imaging sequences is that they all acquire data in a 14sequential fashion. The MR signal is always acquired one point and one line at a time, with each separate line of data requiring a separate application of field gradients and/or RF pulses. Hence, the imaging speed is generally limited by the maximum switching rates compatible with scanner technology and patient safety.

The MR signal produced through the use of RF pulse sequences (vide infra) cannot directly be translated into an image. It is necessary to convert from a frequency representation to a location representation. A digital computer, with sufficient memory and storage, performs necessary operations for this conversion (transformation). There are numerous ramifications of the imaging technique, such as (a) sequential point technique, in which a single volume element is selectively excited and observed at a time; (b) sequential line technique, where an entire line of volume elements is excited and observed simultaneously with increased sensitivity compared to the earlier methods; and (c) sequential plane technique, where the sensitivity is further increased when an entire plane of volume of elements is excited and observed at once. The last of these is most widely used in clinical imaging. The variants of this technique are: the projection reconstruction, Fourier imaging, spin-warp imaging, and rotating frame imaging.

In the past few years, magnetic resonance imaging of flow has blossomed into a powerful technique to visualize blood flow in vessels (Fig. 1.8). This technique called as magnetic resonance angiography (MRA), proved to be a valuable tool for evaluating vascular diseases³⁰. In addition to its ability to image vascular anatomy, new, non-invasive approaches for measuring blood flow have in fact expanded the role of MRI in diagnosing vascular pathology such as, cardiac abnormalities, renal blood flow, etc. In MRA, blood vessels are rendered, not in the usual topographic format of standard MR images, but as a projection over a large volume, similar to a conventional angiogram. No external contrast agent is required because moving protons serve as intrinsic flow markers. MRA enables visualization of normal, laminar blood flow within the vascular system and its disruption due to any pathology such as a thrombus (blood clot). Also, MRI can be of particular use in evaluating vessel patency. Present practice is to use the so-called ‘bright-blood’ technique; i.e. signal from flowing protons is enhanced relative to the stationary protons through the pulse sequence and measurement parameters. The bright blood depicts arteriovenous malformations, aneurysms, stenoses and other diseases. Bright blood MRA methodology can be divided into time-of-flight (TOF) and phase contrast (PC) techniques.

Diffusion-weighted MR imaging (DWI) is used to map as well as assess and/or quantitate the microscopic motion of water protons in tissues^{31, 32}. Measuring molecular diffusion may bring several potential useful new approaches to tissues characterization and functional studies. Microscopic motion of water protons namely randomly directed Brownian motion or diffusion might be used as a potential contrast mechanism. For motion in one direction, the mean translational distance, <L>, that a water molecule would diffuse in time ‘t’ is given by the so-called Einstein equation

where D is the diffusion coefficient of water. The effect of diffusion on the MR images may be understood using a SE sequence in which a pair of bipolar field gradient pulses is included (Fig. 1.9). The purpose of these gradient pulses is to magnetically ‘phase tag’ the diffusing spins, in much the same way a radiotracer is used to tag a molecule in the more classical diffusion studies in physiology³¹. In Figure 1.9, the gradient pulse amplitude and duration are denoted by G and δ, respectively, and the interval between gradient pulses by ∆.

where b is the ‘gradient factor’ characterizing the strength and duration of the applied diffusion gradients and is defined as

MRI can also be used to map changes in brain hemodynamics that correspond to brain functions. The ability to observe precise anatomical structures and also which structures participate in specific functions is due to a technique called ‘functional MRI’ (fMRI). fMRI provides high resolution, noninvasive report of the neural activity detected by a blood oxygen level dependent (BOLD) signal^38–40. It is based on the increase in blood flow to the local vasculature that accompanies neural activity in the brain. These results in a corresponding local reduction in deoxyhemoglobin because the increase in blood flow occurs without an increase of similar magnitude in oxygen extraction⁴¹. Since deoxyhemoglobin is paramagnetic, it alters the T2^*-weighted MR image signal. Thus, deoxyhemoglobin is some times referred to as an endogenous contrast-enhancing agent, and serves as the source of the signal for fMRI. Using an appropriate imaging pulse sequence, human brain functions can be observed without the use of exogenous contrast enhancing agent in many clinical MR scanner (field 1.5 T or greater). Functional activity of the brain determined from the MR signal has confirmed known anatomically distinct processing areas in the visual cortex⁴², the motor cortex^{43, 44}, and Broca area of speech and language related activities⁴⁵. The ability to directly observe different brain functions opens an array of new opportunities to advance our understanding of brain organization, as well as a potential new standard for assessing neurological studies and neurological risk^{46, 47} and the recent advances and applications of fMRI are presented in greater detail in Chapter 9 of this volume.

In general, the MR properties of the tissues and their microenvironment are themselves capable of providing tissue contrast, however, manipulation of the signal intensities by external exogenous contrast agents is necessary in a variety of clinical situations to improve contrast between tissues with similar relaxation times in the normal and diseased states. The mechanism of contrast enhancement is related to several parameters: the class of contrast media (paramagnetic or superparamagnetic), their concentration within the tissue, their compartmentalization in various physiological 20and pathophysiological spaces, and the weighting of the imaging pulse sequence used. Thus, formulations that shortens either the tissue T1 or T2 relaxation time has been devised. The T1 shortening agents are hydrophilic chelated paramagnetic metal ions such as Gd³⁺ and Mn²⁺ that have unpaired electrons and behave as molecular magnetic dipole moments. The fluctuating magnetic fields caused by the tumbling molecular dipole moments alter the T1 of tissue water protons. The paramagnetic ions may be chelated to a wide variety of molecules to reduce toxicity and to retard the free tumbling rates of the metal ion centers thereby enhancing their relaxation effects on the tissue protons being imaged.

Imaging of nuclei other than proton is also possible such as ²³Na and ¹⁹F. Such measurements are exacerbated by the low intrinsic sensitivity of these nuclei relative to proton and the low physiological concentrations relative to water. Sodium MRI studies has demonstrated that the method can be useful for monitoring edema development, cerebral redistribution of sodium following ischemia⁴⁸ and myocardial infarction⁴⁹. The technique has been found to be useful to detect early cartilage 21degeneration as well⁵⁰. Since there is no MR visible fluorine in human body, the ¹⁹F MRI methodology depends on the exogenous addition of a fluorinated material. Perfluorocarbons that label vascular compartments at high concentration have been commonly used, especially in the assessment of vascularity and the kinetics of tissue perfusion⁵¹.

Nuclei that have high natural abundance and sensitivity in living systems are most suitable for in vivo MRS. For these reasons, most of the applications of in vivo MRS have been using ¹H and ³¹P nuclei^{10–12, 14–16}. In living systems, there are two main sources of protons namely water and fat that are used for image formation in MRI. However, the protons of water and fat mask the signals of metabolites that are present in low concentration in tissues. Therefore, the detection of resonances of biochemicals with low concentration require special techniques for suppression of water signal, (vide infra). The aim of in vivo MRS is to obtain a spectrum that arises exclusively from the volume of interest (VOI), with the best achievable sensitivity and with minimum contribution from other regions. A multitude of localization schemes has been proposed; however, the techniques that are in clinical use today are limited. In this section the basics of various localization techniques are presented, for in-depth details the readers may refer to many excellent books^{11, 14, 16} available. The approach taken is to provide a brief overview of the observable metabolites in various organs using the various acquisition methodologies with special reference to proton MRS.

The early localization methods like rotating frame spectroscopy uses B₁ field of the surface coil for localization. Methods like depth resolved spectroscopy (DRESS) use B₀ magnetic field gradients and a frequency selective pulse for localization and the methods that are mostly employed in clinical settings include stimulated acquisition mode (STEAM), point resolved spectroscopy (PRESS) and image guided in vivo spectroscopy (ISIS). These sequences acquire spectra from a single volume (SV). Currently, multivoxel spectroscopic [chemical shift imaging CSI or spectroscopic imaging (SI)] methods are also available which are used to acquire spectra from many voxels simultaneously. Since, most of these methods use MR images to guide the VOI localization and are more often termed as image guided localization methods. The following sections provide a brief survey of some of the early localization methods and those that are commonly used in clinical setting.

Surface coil is a small circular loop of wire that is positioned adjacent (proximal) to a sample for excitation and/or detection and is positioned in such an orientation that a major component of B₁ field generated by the coil is orthogonal to the B₀ field.

Surface coil provides rough localization, i.e. metabolites in the sample close to the coil are detected with good sensitivity, whereas metabolites far from the coil are not detected⁴. In addition by varying the RF pulse length, spatial selectivity can be achieved. The disadvantages include extremely inhomogeneous transverse magnetic field, difficulty in assessing VOI that is below the surface and contamination of signals from extraneous tissues. For these reasons, surface coils are used with other techniques like the use of adiabatic pulses for precise localization and phased array surface coil with correction algorithm for inherent inhomogeneities in the signal reception profile.

DRESS sequence shown in Figure 1.10 was developed by Bottomley et al⁵⁹ is the first spatial localization technique employing magnetic field gradients and was successfully used to acquire localized ³¹P and ¹H spectra from humans and animals. In this method, a single slice is excited by applying a slice selection gradient to the plane of the surface coil with a frequency selective RF excitation pulse to provide depth resolution. The observed volume is the intersection of the excited slice and the sensitive volume of the coil, and is approximately a thick disc. Spatial localization depends critically on the excitation pulse and the surface coil used for detection. In DRESS, the sensitivity reduces as a function of the distance from the cylindrical axis of the surface coil⁵¹. For observation of a deep organ, signals from the surface regions can be suppressed effectively but sensitivity of DRESS sequence is directly related to the depth of the selective pulse that poses a great limitation. Another problem with DRESS is that significant loss of signal occurs for metabolites with short T2 due to short delay of a few milliseconds 24between excitation and detection necessitated by the application of a gradient-refocusing lobe for acquisition.

Localization methods are mostly image guided in that they use proton images to guide the placement of VOI. Hence, MR images are acquired in all the three planes and frequency selective RF pulses are applied during the application of a gradient. Prior to carrying out localized MRS, magnetic field gradient spatially encodes the resonance frequencies and the frequency selective pulse excites the spin distribution within the sample. Localization techniques may be categorized as single volume (or voxel) or multivolume. The position and size of the VOI are determined by the region bounded by the intersection of the three orthogonal planes and the actual shape would normally depend on the slice profiles of the selective pulses. There are different pulse sequences available in the literature and brief details of some of the most commonly used sequences for proton MRS like PRESS and STEAM are described here.

PRESS or double spin echo was proposed by two groups^{59, 60} is an adoption of the 1D DRESS technique to localization in 3D. For proton MRS, initially chemically shifted selective saturation (CHESS) pulses are needed to suppress the water resonance. Spatial localization is achieved by three frequency selective pulses applied in the presence of an orthogonal gradient (Fig. 1.11B). The localized VOI is at the intersection of the three orthogonal slices (Fig. 1.11A). The first 90° pulse is applied in the presence of gradient (e.g. along the x-axis) that selects a slice orthogonal to this axis.

The magnetization of the selected slice is allowed to precess during TE1/4 and is refocused by a second pulse applied in the presence of a gradient along an orthogonal direction (e.g. the y-axis). At TE1/4 it refocuses the magnetization in a slice orthogonal to this axis. Finally, after a free precession delay TE2/4, a second refocusing pulse is applied in the presence of a gradient along the direction orthogonal to the first two directions (z-axis). This last pulse refocuses the transverse magnetization from a slice orthogonal to the first two slices. Therefore the acquired signal results from the volume element common to the three slices, producing a three-dimensional localization in one single scan. PRESS offers the important advantage of a factor of two gain in S/N, although signals are T2-weighted and J-modulated. The VOI definition and localization in PRESS are inferior to other sequences such as STEAM since the slice profiles of the 180° pulses are often worse than those of a 90° pulse. Further, the power requirement for 180° pulses is twice that of a 90° pulse, which may cause RF heating of tissues. Another disadvantage is the lengthened minimum echo time that is an important limitation for metabolites with short T2.

This method shown in Figure 1.12 uses three frequency selective RF pulses that are applied in succession in the presence of mutually orthogonal x, y and z-field gradients^{61, 62}. Each pulse is of 90°, and the second 90° pulse brings the magnetization to the longitudinal direction where it remains for a time period TM. The echo time TE is independent of TM.

In all four echo signals are formed following the last RF pulse: one at TM arises from the second and third pulses, one at (TE/2 – TM) is formed by the first and second pulses, one at (TE/2 + TM) results from the first and third pulses and the stimulated echo is formed at the end of second TE/2. Gradients in the sequence are carefully set such that only the stimulated echo is acquired. All the unwanted echoes and the FID are suppressed using spoiler gradients. To achieve optimal efficiency each pulse should be 90°; however, it is not crucial to localization so long each of them is less than or equal to 90°. This reduces total investigation time and also since pulse angle can be less than 90°, the RF power deposition on tissues is less. However, whatever the nominal pulse angles, STEAM requires a complete dispersion of phases during the first echo delay and destruction of transverse magnetization during TM. The high quality localization obtained with STEAM has made it one of the most frequently utilized single voxel MRS technique despite the disadvantage that there is loss of a factor of two in S/N as compared to Carr-Purcell echoes. The water resonance is suppressed using CHESS pulses at the beginning of the sequence.

Single voxel spectroscopic technique can be extended to multivolume called chemical shift imaging (CSI) or SI^{63, 64}. The spatial information (encoding) is accomplished by switched gradients of the B₀ field.

The data in SI is collected in the absence of a readout gradient (Fig.1.13) to preserve the chemical shift information while in MRI readout gradient is used to obtain spatial information in one of the dimensions. Similar to the phase-encoding in MRI, a sequence of gradient values generates a complete set of spectra which can be spatially and spectroscopically resolved after multidimensional FT. Spectroscopic imaging procedures can be applied for 1D, 2D or even 3D acquisition of an organ. Since, localized spectra from many locations can be acquired simultaneously; SI is a more attractive method than single voxel techniques. The advantages of SI include acquisition of spectral data from small VOIs and the ability to reconstruct low-resolution metabolite images from the spectral data in which the pixel intensity is proportional to the relative concentrations of the metabolites. These metabolic maps or images are useful in visually assessing the spatial variation in the metabolite concentrations. EPI-based SI sequences are now available, which decrease image acquisition times.

In MRI, images are obtained from the strong signals from the inherent high MR sensitive protons which are in high concentration in biological tissue. However, in proton MRS, detection of resonances from metabolites with lower concentrations in the presence of a large water proton signal is a major problem. The concentration of water in tissues is of the order of 70M compared to the millimolar (mM) concentration of other metabolites. Hence, there is the need for water suppression from the VOI and several methods have been developed that exploit the chemical shift differences between the water and the other metabolites^69–71. The common one is to use chemical shift selective (CHESS) RF pulses⁷² which excite a limited narrow band (∼ 60 Hz) of RF frequencies corresponding to the water signal in the sample.

In this method, one or more 90° RF excitation pulses are applied at the water signal and in addition ‘crusher’ gradients are used to eliminate the excited water signal in the transverse plane (Fig. 1.14). The other metabolites are given the RF excitation in the selected region and detected in the period when water magnetization is beginning to recover and not available for excitation. The efficacy of water suppression when using CHESS depends on B₀ and B₁ homogeneities and on T1 of water. B₀ inhomogeneities can be reduced by optimized shimming and similarly B₁ inhomogeneities may be eliminated by using adiabatic RF pulses. Most commonly used techniques of in vivo proton MRS like STEAM and PRESS, use CHESS pulses in the beginning of the sequence for water suppression (Figs. 1.11 and 1.12).

Over the past decade in vivo MRS has increasingly been applied in clinical setup using nuclei like ¹H and ³¹P to study the metabolism of several disease processes^{11, 15, 16} (see Chapters 11, 13–16). MRS can be carried out using different nuclei that include ¹H, ³¹P, carbon (¹³C), lithium (⁷Li) and fluorine (¹⁹F). Using MRS it is possible to detect certain biochemicals that participate in different metabolic pathways. It provides information about energy status [phosphocreatine (PCr), inorganic phosphate (Pi) and nucleotide phosphates (NTP)], phospholipid metabolites [phosphomonoesters (PME), phosphodiesters (PDE)], intracellular pH and free cellular magnesium concentration through ³¹P MRS. Natural abundance or enriched ¹³C MRS provides information about glycolysis, gluconeogenesis, amino acids and lipids. Water-suppressed ¹H MRS shows total choline (TCho), total creatine (TCr), lipids (Lip), glutamate (Glu), glutamine (Gln), inositols (Ins), lactate (Lac) and in brain, N-acetyl aspartate (NAA). ⁷Li and ¹⁹F MRS can be used to study pharmacokinetics and their applications are outlined in Chapter 18.

Biomedical MR has achieved amazing level of success as an important tool through MRI and in vivo MRS to study several disease processes. MRI is a multiplanar imaging capability, has high spatial resolution, excellent soft tissue contrast, and uses no ionizing radiation. It covers a broad range of applications from fast noninvasive anatomical measurements, to studies on tissue physiology and metabolism. Since it is noninvasive, studying chronic diseases has several advantages for monitoring disease progression and therapy response. MRI is also useful for pharmacodynamic studies in humans to assess whether new compounds are exerting the desired pharmacological effects. Such studies will serve and guide the dose selection for the extensive and expensive clinical efficacy trials. MRI has also been developed as a powerful tool to guide interventional procedures in clinical medicine. Designs of new open magnet systems allow improved patient access for MR-guided interventional procedures with the use of MR-compatible needles and catheters. Applications of this interventional MRI include aspiration cytology, chemoablation, cryoblation, etc. As MR technology matures, the technological limits on the achievable scanning speed and pixel resolution are rapidly approaching the limits imposed by the laws of physics.