Modalities and methods
Magnetic resonance imaging
Magnetic resonance imaging, MRI, is the latest newcomer of the radiological modalities. MR units can provide sectional images in any plane of any part of the body. No ionising radiation is involved, and air or bone, represent no obstacle to imaging. Compared to ultrasonography and computed tomography, the modality is more expensive, technically more advanced, and theoretically more difficult to comprehend. Nevertheless, MRI has completely revolutionised some areas of diagnostic radiology. The usefulness of MRI will be reviewed in the organ-related clinical chapters; the present chapter will concentrate on giving a simplified explanation of the most basic principles of this new imaging technique.
The MR unit and its magnetic field
The most basic components of a MR unit are a very strong magnet, a radio transmitter, a radio frequency receiver coil, and, of course, a computer. The interior of the magnet is often tunnel-shaped and big enough to contain a human adult (Fig. 15). Most magnets have a magnetic field orientated parallel to the long axis of the patient (Fig. 16).
Patient preparation prior to MR imaging. Before start of scanning, the patient table is elevated and fed into the tunnel-shaped interior of the magnet. The anatomic region of interest must be located in the centre of the magnet. The entire procedure may last for approx. 1/2-1 hour, and during this period, the patient must remain nearly motionless (sedation is needed for small children). (Phototechnical Department, Rikshospitalet, Oslo.)
The MR magnet. Most magnets are electromagnets with a horizontal magnetic field (B0). During imaging, the patient lies in the tunnel-shaped interior of the magnet. The z, x, and y co-ordinates are shown. Permanent magnets with vertical magnetic fields are available as well.
The magnetic field of the strong magnet is designated B0, and is illustrated as a vector, i.e., an arrow whose orientation shows the direction of the magnetic field from south to north, and whose length indicates the strength of the magnetic field. The orientations within the magnet are shown by means of an imaginary frame of reference with three co-ordinates, z, x, and y (Fig. 16). The z-direction is always the direction of the magnetic field, B0, and when this field is parallel to the long axis of the patient, the horizontal axis perpendicular to z is named x, and the vertical axis is named y. The plane through x and y (the x-y plane) is thus orientated perpendicular to the magnetic field, B0. The strength of the magnetic field is measured in tesla (T) or gauss, where l tesla = 104 gauss. For clinical MR imaging, field strengths from 0.02 tesla to 2.0 tesla have been used (experimentally 4.0 tesla has also been used). Most MR units have field strengths from 0.1-1.5 tesla. For comparison, the earth has a magnetic field strength of 0.7 gauss at the poles and 0.3 gauss at the equator.
Hydrogen nuclei (protons) in a magnetic field
Magnetic resonance imaging exploits the fact that hydrogen nuclei - in this context often named protons - are tiny magnetic dipoles with a north pole and a south pole. The magnetic moment of one proton is often termed (Fig. 17). When a patient is placed within the strong magnetic field of an MR magnet, all the tiny proton magnets of the body line up in the direction of the external field (B0) (similar to compass needles adjusting to the magnetic field of the earth). In addition, the magnetic axis of each proton starts to rotate around the direction of the external magnetic field (Fig. 17). This particular rotational motion is called precession, and its frequency is named the resonance frequency or Larmor frequency (after the French physicist Larmor). The Larmor frequency (w0) is proportional to the external magnetic field (B0):
w0 = g · B0
The equation is called the Larmor equation, and g is a constant named the gyromagnetic ratio. This ratio (w0/B0) is specific to each type of magnetic atomic nucleus, and for the hydrogen nucleus, the ratio is equal to 42.58 MHz/tesla. This means that at the magnetic field strengths used in MRI, the Larmor frequency of the hydrogen nucleus is in the radio frequency range (42.58 MHz at 1.0 tesla).
Precession. The magnetic moment of one proton is illustrated as a vector (
). The vector indicates the direction of the proton magnetic field from south to north (the magnetic axis). In a strong, external magnetic field (B0), the magnetic axis of the proton will rotate (precess) around the B0 (z) direction, the north pole (and south pole, not shown) describing a cone-shaped figure. (The circle in the origo of the frame of reference indicates the proton.) (
0: the Larmor frequency.
Unlike ordinary compass needles, the magnetic moments of the precessing protons do not all point towards the north pole. A small majority of protons precess with their magnetic moments aiming towards "north", i.e., in a direction closely parallel to the external magnetic field. These are consequently called "parallel protons". The remainder of the protons precess with their magnetic moments aiming towards "south", i.e., close to anti-parallel to the external magnetic field. These are the "anti-parallel protons". The result is the creation of a net magnetic moment in the tissues of the patient; the tissues become magnetic, and the magnetism (M) is oriented exactly parallel to the external magnetic field, B0. At first, the tissue magnetism has no precessional motion. Although the individual protons all precess, they are evenly distributed around the B0 direction, leaving no magnetic component in the x-y plane (Fig. 18 a). The size of M is determined by the surplus of parallel protons. The surplus is proportional to the external magnetic field strength, but is always very small, only in the order of 1-10 protons per 1 million protons. M is also proportional to the number of protons per volume unit of tissue, i.e., the x proton density. The enormous amount Of protons (i.e., hydrogen nuclei) present in most tissues (approximately 1022 per ml water), accounts for the fact that the net magnetic moment, M, is strong enough to induce an electric current in a receiver coil located outside the patient. These induced "MR signals", as will be explained below,Are used for the reconstruction of MR images.
a) The tissue magnetism (M) is created by a surplus number of "parallel" protons (see text). The magnetic vectors of the individual protons (thin arrows) are evenly distributed around the z-axis, and M is therefore oriented exactly in the z-direction.
b) A 30° radio frequency pulse has rotated all x protons and M 30° away from z, in the clockwise direction. Due to never ending proton precession around the z-axis, M is also precessing around z.
c) The result of a 90o pulse: M is precessing around z in the x-y-plane.
The MR signal
Any magnetism may induce an electric current in a coil, but a prerequisite for this to happen, is a change in the magnetic field strength running through the bore of the coil. To make the tissue magnetism, M, induce a current in a coil, radio waves are needed. Radio waves are electromagnetic waves, containing both an electric and a magnetic field. When a short electromagnetic radio frequency pulse is transmitted into the patient along the y-axis, the magnetic field of the radio waves will force the magnetic moments of all the protons to rotate in a clock-wise direction around the y-axis. For this to happen, the frequency of the radio waves must be exactly equal to the Larmor frequency of the protons. This is the phenomenon termed magnetic resonance.. Resonance means synchronous vibration, and in this context it implies that the magnetic fields of the protons and the radio waves must resonate, i.e., have the same frequency, in order to change the orientation of the proton magnetic moments.
a) After the transmission of a 90o pulse, the tissue magnetism (M) is inducing an electric current (MR signal) in the receiver colt. The signal strength determines the shade of grey of the corresponding area in the final image.
b) The situation in a) may be compared to a rotating bar magnet (= the tissue magnetism M) inducing an electric current in a colt being connected to a light bulb. The amplitude of the current determines the light intensity of the bulb (= shade of grey in the MR image).
When the surplus parallel protons are rotated away from the B0 direction, M must follow (Fig. 18 b). The protons will continue to precess around the z-axis (they are forced to do so by the B0 magnetic field), and M will consequently also start to precess around the z-axis (Fig. 18 b). The strength and duration of the radio frequency pulse determine how many degrees M is rotated away from the B0 direction, and the pulse is named accordingly. The result of a 90° pulse is thus that M (for a short period of time) will rotate in the x-y plane, perpendicular to the B0 direction (Fig. 18 c).
A receiver coil is placed on the outside of the anatomical region with its bore oriented towards the patient, perpendicular to the B0 direction. When M rotates in the x-y plane, it will induce an electric current in the coil, and this electric current is called the MR signal (Fig. 19 a). These (or similar) signals are used for reconstruction of sectional MR images. The situation after a 90° pulse is analogous to a bar magnet rotating past the opening of a coil (Fig. 19 b). The varying magnetic field through the coil will induce an electric current, and if the coil is connected to a light bulb, the bulb will shine. The stronger the magnet, the brighter the light. The same principle applies to MR imaging; tissues exhibiting a strong magnetism (M) will induce strong signals and appear bright in the image, and tissues exhibiting a weak magnetism will induce weak signals and appear dark.
Image contrast: Proton density, T1-, and T2-weighting
As explained above, contrast in MR images is determined by differences in tissue magnetisms, or more precisely, by the different strengths of magnetism that rotate in the x-y plane and induce currents in the receiver coil. Tissue magnetism is first of all determined by the proton density (see above). Anatomic areas containing very few protons, like air, will always induce very weak MR signals and therefore always appear dark in images. Water and other fluids, on the other hand, having a very high proton density, presumably should always appear bright in MR images. This is not the case, however. Depending upon to the method used for image acquisition, fluids (like CSF) may appear either bright or dark. The reason for this perhaps confusing fact is that proton density is not the sole determinant of image contrast. Several other parameters play a role as well, and the two most important of these are named T1 and T2.
To reconstruct an image, several MR signals are needed, and several radio frequency pulses must therefore be transmitted. Between the pulse transmissions, the protons undergo two different relaxation processes, T1- and T2-relaxation. The rapid decay of the induced signal seen in Figure 19 a, is partly the result of T2-relaxation. The decay is a consequence of the gradual disappearance of the magnetism in the x-y plane, Mxy, caused by small differences in the local magnetic field strength (partly due to magnetic molecules in the tissues). Protons exposed to slightly different magnetic field strengths will have slightly different Larmor frequencies, and the surplus "parallel" protons that were closely packed around Mxy immediately after the 90° pulse (Fig. 18 c), will dephase, i.e., spread out around the z-axis. When the individual protons are evenly distributed around the z-axis, Mxy is gone. This loss of net magnetism in the x-y plane is called T2 relaxation, and T2 is defined as the time until Mxy has lost 63% of its original, maximum value. A common T2 value in parenchymal tissue is approximately 50 ms. After a time equal to 4-5 times the T2 value, Mxy will have disappeared completely. The T2 value varies considerably, however, with the physical and chemical properties of the tissues. Fluid and fluid-like tissues typically have a long T2 (Mxy and the MR signal disappear slowly), and solid tissues and substances have a short T2 (Mxy and the MR signal disappear rapidly).
T1 relaxation is a slower process than T2 relaxation, and involves the gradual alignment of the individual protons with the B0 direction, thus restoring the situation prior to the 90° pulse (Fig. 18 a). During this process, the net magnetic moment along the z-axis, Mz, will increase from zero with ever decreasing speed until its maximum value, determined by the proton density in the tissue, is reached. T1 is defined as the time until Mz has regained 63 % of its original, maximum value. The shorter the T1, the faster the restoration of Mz, After a time equal to 4-5 times the T1 value, Mz is completely regained. A common T1 value in parenchymal tissue is approximately 500 ms. There is, however, a large variation in T1 in the different tissues. The T1 value is largely determined by molecular size and mobility. Generally, T1 is shortest in tissues having molecules of medium size and mobility, e.g. adipose tissue. Smaller, more mobile molecules (as in fluids) and larger, less mobile molecules (as in solids) have longer TI values.
T1-weighting. T1 relaxation curves showing how the magnetism in the z-direction (M) in two different tissues (A and B) increase from zero after repetitive 90° pulses. The shaded parts of the first two relaxation curves indicate how Mz would have increased until maximum if the next 90° pulse had not been transmitted. The Mzs of tissue A and B would have levelled out at the same maximum value, indicating similar proton densities in the two tissues. The repetition time (TR) is so short, however, that T1 relaxation is not completed when the next pulse is transmitted. At pulse transmission, tissue A, having the shortest T1, will have regained a larger Mz than tissue B, and tissue A will therefore induce a stronger signal in the receiver coil after each 90° pulse. The difference in signal strength is caused by differences in T1, hence the term T1-weighted image.
Proton density and T2-weighting. The regain of Mz and loss of Mxy are shown for two tissues (A and E). having short T1 and T2 (A), and long T1 and T2 (E), respectively. After having recorded the MR signals, the waiting time until the next 90o pulse is sufficiently long to complete T1 relaxation in both tissues. The possible effect on contrast by differences in T1, is thereby eliminated. Early signal registration provides proton density weighting (PD); late registration gives T2-weighting (T2).
By adjusting the time period between the radio frequency pulses transmitted, the operator of a MR unit may decide whether image contrast should be determined mainly by proton density, T1 or T2. A certain time interval between the pulses is needed to allow regaining of Mz. The longer the time interval (up to a certain point), the larger the Mz to be rotated into the xy-plane by the next 90° pulse, and the stronger the MR signal induced. If the next 90° pulse is transmitted before completion of T1 relaxation in the tissues, the size of Mz in the tissues will depend upon their T1 values. Provided fairly similar proton densities, tissues having the shortest T1 will have regained the largest Mz, and will consequently induce the strongest MR signals after the following 90° pulse (Fig. 20). These tissues will therefore appear bright in the final image. Tissues with the longest T1 will similarly induce the weakest signals. MR images where contrast is largely determined by differences in T1, are called T1-weighted images. The time interval between the radio frequency pulses is named repetition time (TR), and T1-weighted images are acquired with relatively short TRs (approximately 500 ms).
By increasing the TR, it is possible to achieve alternative image contrast; either proton density weighted images, or T2-weighted images. In proton density (PD) weighted images, the tissues with the highest proton density induce the strongest MR signals and appear bright, and in T2-weighted images, the brightest tissues are those having the longest T2. For both types of contrast long TRs are needed to eliminate the effect of differences in T1 on image contrast. Using TR values mare than 5 times the longest T1 of the tissues, ensures that all tissues have completely regained their maximum Mz before transmission of the next 90° pulse (Fig. 21). The maximum magnetism along the z-axis is determined by the proton dens it y, and the relative strengths of the MR signals derived from the various tissues immediately after the 90° pulses, will therefore reflect the relative proton densities of the tissues. The contrast becomes PD-weighted. T2-weighted contrast is achieved by introducing a time interval (called echo time, TE) between the 90° pulse and the signal measurement. During this time interval, the size of Mxy is gradually reduced due to T2 relaxation; slowly in tissues having a long T2, more rapidly in tissues having a short T2. The amplitudes of the induced MR signals recorded at the end of the echo time, will therefore reflect the differences in T2 in the tissues (Fig. 21). (A detailed explanation of how the MR signal- in this context termed echo - is actually measured at the end of the echo time, is beyond the scope of this chapter. The interested reader should consult more in-depth literature.)
It should be clear from the above that image contrast in MR imaging can be made much more variable than image contrast in alternative modalities such as computed tomography and ultrasonography. Image contrast is determined by operator-dependent parameters such as repetition time and echo time, and by tissue-dependent parameters such as proton density, T1, and T2. A basic knowledge of these parameters is necessary for proper evaluation of MR images.
Slice selection. Magnetic field gradients
Radio frequency pulses will cause induction of MR signals only when the pulse frequency is exactly equal to the proton Larmor frequency. This fact makes it possible to collect MR signals from a predetermined thin slice of tissue. The first step towards slice selection is to create a magnetic field gradient through the anatomical region to be imaged (Fig. 22). Special coils (gradient coils) create small additional magnetic fields with the result that the strength of the B0 magnetic field increases linearly in one direction. The Larmor frequency of the protons is proportional to the magnetic field strength (see the Larmor equation), and the Larmor frequency will therefore increase linearly in the gradient direction. By transmitting radio frequency pulses having a predetermined narrow frequency range, MR signals will be recorded from only the thin slice of tissue that has a Larmor frequency range corresponding to the pulse frequency range. The orientation of the magnetic field gradients and therefore also the slice directions, are freely selectable.
Reconstruction of the MR image
The tissue slice to be imaged, may be considered as consisting of several equally large volume elements, voxels (see Fig. 7 in the section on CT). After each 90° pulse, every voxel has its own tissue magnetism (Mxy) which induces a signal in the receiver coil. The amplitude of the individual voxel signal is determined by the size of the voxel tissue magnetism, which again is determined by voxel dependent factors such as proton density, T1, and T2, and choice of repetition time and echo time. Each voxel corresponds to a picture element, pixel, in the final two-dimensional image. The brightness (shade of grey) of the pixel is determined by the signal amplitude induced by the magnetism in the corresponding voxel.
The MR signal recorded from a slice of tissue, is a composite signal induced by all the individual voxel magnetisms simultaneously. The MR computer needs to differentiate between the various voxel signals to assign the correct brightness to each pixel. To enable the computer to do so, each voxel signal must be given a unique and recognisable code. The code being used, is the frequency and phase of the voxel signal, which is determined by the frequency and phase of the rotating voxel magnetism (Mxy), The encoding is done by two
Slice selection (axial slice). By means of gradient coils, the external magnetic field increases linearly in the z-direction. Planes per pendicular to this magnetic field gradient have uniform magnetic field strength. In the axial plane at the field strength B0, the Larmor frequency of the protons is
0 cranial to this plane, field strengths and Larmor frequencies are higher, caudal to this plane they are lower. If a 90o pulse having the frequency w0 is transmitted, only the protons in the plane having the exact same Larmor frequency will be rotated 90 and induce a signal.
Phase and frequency encoding. Part of a tissue slice in the x-y plane (axial slice) is shown with voxels. A brief magnetic field gradient in the y-direction applied between pulse transmission and signal registration, provides the voxel signals with phases determined by the voxel locations in the y-direction. A magnetic field gradient in the x-direction applied during measurement of the MR signal, provides the voxel signals with frequencies determined by the voxel locations in the x-direction.
magnetic field gradients, applied in the y- and x-direction, respectively (for axial slices). The gradients affect the rotation of the voxel magnetisms in such away that the resulting voxel signals are given a phase determined by the voxel location along the y-axis, and a frequency determined by the voxel location along the x-axis (Fig. 23). As a result, each voxel is given a unique code of phase and frequency. (A detailed explanation is beyond the scope of this chapter.)
To extract the various frequencies and phases contained in the composite MR signal, a complicated mathematical analysis termed two-dimensional Fourier transform is used. This method is dependent upon the information contained in numerous repetitive signals derived from the same slice, and this is the reason why "conventional" MR imaging is relatively time-consuming.
Circulating blood: a natural "contrast medium"
In MR imaging, stagnant blood has a signal intensity (brightness) determined by the proton density, T1, and T2 of blood, and by the contrast "weighting" chosen. (In practice, stagnant blood appears bright at most "weightings".) Circulating blood, on the other hand, will - due to its flow velocity - in most instances induce no MR signal, and thus behave as an effective "negative" contrast medium. Vessel lumina and heart chambers will appear dark and clearly delineated by the brighter surrounding stationary tissue.
There are, however, special MR techniques that make circulating blood appear bright and stationary tissue appear dark. These techniques are being used in MR angiography (MRA) to create two-dimensional projection images of three-dimensional vascular structures. After a single image acquisition, the vascular anatomy may be viewed from any angle.
MR contrast media
Five to ten years ago, contrast media for MR imaging were considered completely unnecessary. In many clinical situations this is still true. Experience has shown, however, that contrast media may increase the diagnostic information in several disease states. During recent years, a growing number of contrast media have been developed. They all have magnetic properties, and they change the signal intensity of the tissues where they are located, by shortening the relaxation processes (T 1 and/or T2) of the surrounding protons. The most commonly used contrast media contain the paramagnetic metal ion gadolinium (Gd3+), chelated to a carrier molecule. These contrast media are administered by intravenous injection, and they have a distribution in the body similar to water soluble X-ray contrast media.
Contraindications and potential dangers
No harmful effects of the static or fluctuating magnetic fields used in MR imaging, have been shown. Ferromagnetic objects are subjected to very strong mechanical forces, however, and any ferromagnetic object having a location where motion of the object may be harmful to the patient, represents an absolute contraindication to MR imaging. The most important and dangerous objects are ferromagnetic intracranial aneurysmal clips and intraocular ferromagnetic foreign objects. The main potential danger involved with these objects is serious haemorrhage. The presence of a pacemaker represents an absolute contraindication to MR imaging. The function of the pacemaker may be affected by the magnetic field, and furthermore, electric currents may be induced in the pacemaker electrode with possible heating of the endocardium.
The transmitted radio frequency waves will always have a heating effect on the tissues. To avoid harmful heating, the maximum allowable energy transmitted to the patient is regulated by international recommendations. First trimester pregnancy is by some considered an absolute contraindication to MR imaging due to possible heating of the foetus. In the first trimester, the foetus is surrounded by a relatively large amount of amniotic fluid, and has little capability to remove the extra heat.
Magnetic resonance spectroscopy
MR units having a magnetic field strength of at least 1.5 tesla, may also provide the possibility of undertaking in vivo magnetic resonance spectroscopy (MRS). In vitro MRS has been known since the 1940s, and is a much older technique than MRI. MRS is based upon the fact that magnetic atomic nuclei and molecules located in a magnetic field will cause local changes in the field strength depending upon the molecular structure and composition. Magnetic atomic nuclei of the same type (e.g. hydrogen) will therefore have Larmor frequencies that vary slightly with the molecular locations of the nuclei. The MR signal induced after a radio frequency pulse will contain these frequencies. A frequency analysis of the composite MR signal generates a frequency spectrum, which is an amplitude versus frequency display showing the different frequencies present and their respective amplitudes. Such a frequency spectrum can tell the presence and relative concentration of numerous molecules or metabolites.
Several magnetic atomic nuclei may be used in MRS, but the two most commonly used nuclei in in viva MRS are hydrogen (1H) and phosphorous (31P). There are large differences in Larmor frequency between different types of nuclei, and the receiver coil may receive signals from only one type of nucleus at a time. MR imaging and MR spectroscopy may be combined, e.g. by first performing (proton) MR imaging for localising purposes. An area of interest may be selected from the MR images before switching to a phosphorous receiver coil for phosphorous spectroscopy. The result may be displayed as a frequency spectrum, or may also be shown as colour coding of areas in the grey scale MR image, the colours indicating the location and concentration of various phosphorous compounds such as A TP, ADP, or inorganic phosphate. Hydrogen (proton) spectroscopy may also be done, and local concentrations of e.g. lactate indicative of ischemia may be shown. In viva MRS thus makes it possible to acquire information on important metabolic processes in both normal and pathological tissue, and to follow functional effects of treatment. The great potential of in viva MRS has been realised for many years, but the method still awaits its breakthrough in everyday clinical practice. The technical difficulties have been more profound than expected, and we probably have to wait still some years until in viva MRS becomes a natural part of the diagnostic armamentarium of radiology.