INTRODUCTION TO M.R.I. HARDWARE

GREG BROWN
Senior Radiographer MRI
Royal Adelaide Hospital
These notes outline the major components of a Magnetic Resonance Imaging (MRI) machine. The intention is to familiarise the reader with the primary functional components common to all MRI machines by discussing design variations, key specifications, and the role of each component in the process of image formation.

Image Creation & Functional Overview
Primary System Components
Magnet
Magnet Types
Magnetic Shielding
Shim System
Passive Shim
Resistive Shim
Gradient Shim
Gradient System
RF System
Image Processor
Console & Workstations
Host Computer
Patient Communication and Monitoring
Image Output Devices


THE PROCESS OF MR IMAGE CREATION

When a patient is placed in a strong magnetic field their tissue becomes magnetised primarily due to the behaviour of certain magnetic nuclei. This nuclear magnetic field has a component which rotates at a specific frequency due to a phenomenon known as nuclear magnetic resonance (NMR). The resonant frequency is proportional to the strength of the applied magnetic field, and known as the Larmor Frequency
Virtually all clinical MR imaging relies of manipulating the nuclear magnetic field of hydrogen which has a Larmor frequency of 63.87 Mhz. at a field strength of 1.5 Tesla. The MRI image is a digital map of NMR signal intensity. To arrive at a useful image, the MRI scanner must create detectable NMR signals from tissues so that different tissues will give different signal intensity (contrast), and spatially localise the NMR signals.
 


FUNCTIONAL OVERVIEW OF THE MRI SYSTEM


MAGNETS

The magnet of an MRI scanner provides a stable, strong and spatially uniform magnetic field within a structure that allows adequate patient access.
The unit of magnetic field strength is the Tesla (T). Another, less frequently used, unit of magnetic field strength is the Gauss (G).     1 Gauss = 10-4 Tesla 1 Tesla = 10,000 Gauss.
MRI scanners use a magnetic field strength of 0.04 T to 2 T with most in the 0.5T - 1.5T range. By comparison, the earth's magnetic field is about 0.5 x 10-4 T, and a fridge magnet is about 0.2 Tesla.
Three types of magnets are available:-
  • Permanent
  • Resistive
  • Superconductive
  • Permanent Magnets

    Permanent magnets are constructed from magnetised ferromagnetic material such as alloys of iron and cobalt (ALNICO), or more efficiently with rare earth alloys such as Neodymium Iron Boron (Ne-Fe-B) or Samarium Cobalt (Sm-Co) ALNICO permanent magnets are extremely heavy considering the achievable field strengths. A 0.2 T magnet using ALNICO weighs 23 Tonnes, using Ne-Fe-B the weight is reduced to 4 tonne unit, but the material is very expensive. Virtually all permanent magnets use a vertical main field alignment with the patient sliding in horizontally between two close poles. This geometry require different RF coil and gradient coil designs to that used in horizontal field magnets. The field of a permanent magnet is always "on", yet they require no applied power and have a stable field as long as the room temperature is well controlled. Field homogeneity is relatively poor compared to that of superconductive magnets, and is generally achieved over a smaller volume.
    Older commercial MR systems have offered permanent magnets between 0.064 T and 0.3T but they constitute a very small portion of the installed base. More recently GE and Hitachi released systems based on a 0.2 Tesla permanent magnet. Marketing of these scanners concentrates on their potential in interventional, biopsy, and joint motion work as well as helping to claustrophobia in some patients.

    Resistive Magnets

    In a resistive electromagnet the field is generated by the flow of current through resistive electrical windings. Resistive systems require the constant application of large amounts of electrical power to maintain the field. A 0.15T magnet uses 50 kilowatts, and a 1.5T magnet would theoretically consume 5 Megawatts. Most of this power is dissipated as heat necessitating elaborate cooling systems.
    Air core resistive magnets can be designed with 2 or 4 sets of solenoidal windings arranged co-linearly around the magnet bore to yield a horizontal magnetic field. Iron core resistive designs generally use two solenoid windings around the limbs of an iron rectangular yoke fitted with large poles in the horizontal limbs and yielding a vertical field at the patient bore. The yoke can augment the field of the electromagnetic windings by up to 40%.
    Resistive windings can also be used to augment the fields of permanent magnets in a hybrid configuration that offers less mass than the permanent approach and less power usage than the resistive approach for a given field strength, although the heat produced by the resistive windings and the temperature dependence of the permanent magnetic material can create design difficulties.
    Field homogeneity is relatively poor compared to that of superconductive magnets, and is generally achieved over a smaller volume restricting high quality imaging with gradient echo sequences and large fields of view.
    General purpose scanners using resistive magnets have been offered at between 0.1T and 0.3T but occupy less than 1% of the world-wide installed base of about 5000 scanners. This figure has increased with the release by Siemens in 1993 of the Magnetom OPEN, an interventional MRI unit based on a 0.2 Tesla resistive magnet by Oxford Magnet technology. Picker based its OUTLOOK system on the same magnet but runs at slightly higher field strength. Other new resistive designs are aimed at the niche market of the small parts scanners, where they offer suitable field strengths at low cost for the user with a low patient throughput.

    Superconductive Magnets

    The most common MR magnet type is an air core electromagnet made with windings made of superconductive material. The superconductive windings show virtually zero electrical resistance when they are cooled to very low temperatures and so, as long as the material is kept below its critical temperature, the current flow producing the magnetic field continues without application of electrical power.
    MR magnets use multi-filament windings of Niobium-Titanium alloy (Nb-Ti) embedded in a copper core. Nb-Ti becomes superconductive at about 10oK (-263o C.), so the windings are bathed in liquid Helium (4oK) within a structure called the cryostat. The cryostat is constructed of low thermally conductive material and a series of refrigerated zones and extreme vacuums to thermally shield the cryogenic zones from the relatively high temperature outside environment. Liquid Helium (LHe), and in older units liquid Nitrogen (LN2), is constantly boiling away and the cryogen supply needs to be topped up regularly. This is expensive and time consuming. High cryogen usage was a feature of older superconductive magnets, but designs from the past 3 years use significantly less liquid Helium (approx 0.1 litres per hour), and no liquid Nitrogen.

    Once energised the magnetic field is always "on", but it can be turned off (quenched) in about 30 minutes by discharging the current flow through a resistive shunt made by heating a small section of superconductive material in parallel with the main windings. In older designs section this process consumes a large percentage of the cryogens and take a day to restore a stable field (ramping), but new magnet designs effect a controlled quench with little cryogen loss, and can be ramped up within a few hours.

    General purpose scanners use superconductive magnets are available from a variety of manufacturers, operating at fields of 0.5 T to 2 T. Over 90% of the world's installed MRI systems.


    Magnetic Field Shileding

    The magnetic fields generated by high strength superconductive magnets extend over a large area, presenting a number of siting difficulties. This "fringe field" can disrupt the operation of computer storage media, colour video monitors, CT scanners, and Gamma cameras. Dynamic intrusions of ferro-magnetic material into the fringe field will distort the path of the magnetic field lines (flux lines), induce current variations in the superconductive windings, and create reciprocal field distortions within the magnet bore that degrade homogeneity. These transient field distortions will oscillate in intensity, taking as long as 20 minutes to decay effectively. Siting of high field MRI scanners is made more practical if the magnetic field can be shielded or contained to reduce the fringe field area.

    Passive Shielding

    This approach uses ferromagnetic material (usually iron) to create a structure that will contain the magnetic flux lines more effectively than air. The design can place iron close to the magnet, restricting fringe field intensity within the scan room, or incorporate the shield in the walls of the room to contain it at that point. The closer the shield is to the magnet, the more critical is its design, material, and symmetry. In a derivation of high energy linear accelerator magnet designs, the shield can also be design to augment the field strength. Wether the shield is placed close to the magnet or in the walls of the scan room, the mass of a given iron alloy required to achieve a specific degree of fringe field containment is constant. About 20 tonnes of iron will effectively shield magnets of 1 to 1.5 Tesla, sufficient to require careful design of the magnet room floor. By containing the flux lines in the iron the fringe field is greatly contracted, so passive shielded magnets are less able to be affected by moving ferromagnetic objects outside the scan room. Close passive shields, like the Siemens Magnashield are waning in popularity due to their mass and construction expense, although Toshiba uses the technique to achieve part of the shielding offered by its Gemini Shield magnets. Room wall shielding is still popular, particularly for screening high traffic areas, but active shielded magnets have become the norm for most manufacturers.

    Active Shielding

    In this design approach the fringe field is suppressed, rather than contained. Current flow in a set of extra superconductive windings generate a field that oppose a proportion of the field produced by the main windings in a manner that balances out the dipole, or ovoid, pattern of the fringe field, leaving the quadrupole fringe field pattern. The quadrupole field intensity decreases much more rapidly with distance form the magnet centre. (Dipole B µ 1/d3   Quadrupole B µ 1/d5).

    The main windings of a 1 T active magnet generates a 1.5 T field opposed by a 0.5 T shield winding, thus active shield magnets are considerably more expensive than unshielded models. This extra expense is usually offset by avoiding building modifications and the cost of passive shields. Active shield magnets weigh only a little more than unshielded magnets.

    The prime disadvantage of simple active shield magnets is that at high fields they can be more sensitive to moving ferrous objects than an unshielded magnet. This sensitivity results from the dynamic nature of the balance of the shielding field and the main field and their interaction in the centre of the magnet bore to create a homogeneous field for imaging. Transient disturbances in the ferromagnetic environment of the fringe field alter the current flow in the main magnet windings and the shield magnet windings differently, disturbing the balance between their fields and varying their interaction in the imaging region, thus creating a larger inhomogeneity than would have been the case for an unshielded magnet. The creation of a significant field disturbance depends on the magnetic susceptibility of the object, its mass, and the distance form the magnet centre (or more accurately the strength of the fringe field it encounters). Thus small steel objects such as beds, and wheelchairs close to the magnet can create transient problems of field inhomogeneity as well as heavy trucks or traffic much further away.

    Oxford Magnet Technologies has addressed this problem in its recent shielded magnets by adding extra superconductive windings to the outside of the main and shield windings. These "B0 Shield" windings are shorted together but isolated from the magnet windings. In the normal state do have no current flowing. They are constructed so that when the fringe field is distorted the extra currents induced flow in the B0 shield windings rather than the main or shield windings, thus preserving their balance. The current in the B0 shield is discharge automatically at a time when the magnet is not in use. This style of magnet is supplied to Siemens, Picker, and Toshiba for a variety of field strengths using the commercial name of external interference shield (EIS).


    Does Field Strength Equal System Capability ?

    Considerable debate has flourished about the optimum field strength for MRI scanners. Selection discussions which centre purely on the issue of field strength are frequently ideologically or commercially motivated, or indicate a poor understanding of the role other components of an MRI scanner. Field strength will dictate to a certain extent the level of imaging speed, image quality, and capability of the scanner, however the combination of magnet homogeneity, gradient system performance, scan sequences, and coil inventory play a greater role in determining the overall capacity of the machine. The required capability of a unit should be determined relative to its planned use, with potential purchase choices rated against this. The final significant determinant of imaging performance remains the competence of the MRI Radiographer, as displayed through appropriate patient management and technique selection, working to extract the optimum image and examination in each circumstance.

    SHIM SYSTEM
    The spatial uniformity of magnets are designed as near to perfect as possible, but manufacturing tolerances and ferro-magnetic material near the installation site will introduce significant spatial variations in the magnetic field of the sensitive volume that will make imaging impossible or sub-optimal. In addition to these "permanent" influences on field homogeneity, the ferromagnetic environment of a site may change as buildings, car-parks, or structures surrounding the MRI installation are altered through the life of the equipment. These variations must be compensated for by a process called "shimming". Shimming has the aim of creating a large volume highly uniform magnetic field. The resultant field is called the sensitive volume, and is defined as field variations in parts per million, over a specific volume diameter ie. +/- 2 ppm 50 cm. DSV. On a shorter time scale, ferromagnetic patient trolleys, large vehicle traffic around the scan site, and the individual patient, can introduce transient field homogeneity variations.
    Shimming levels are not as critical for spin echo imaging where 180o pulses recall the echo in a manner that compensates for up to 50 ppm of local field homogeneity, but for gradient echo and multi-shot sequences (EPI, GRASE, segmented GE) a high order of field homogeneity ensures a long T2* and the preservation of the Mxy for as long as possible. Fat saturation techniques, offset field of view imaging of joints, particularly with gradient echo sequences, and long field of view spinal imaging with phased array coils also place extra value on the production of a large well shimmed volume. These sequences are more frequently being used in standard protocols, so the homogeneity figures of the magnet and the capability of the shim system have become more critical issues in system design. The uniform field achieved by shimming also has a role in ensuring accurate slice thickness and profile, slice location, and geometric fidelity for all imaging.
    Shimming is typically achieved by some combination of four methods:-

     Passive Shimming

    Small ferromagnetic plates are fitted within the energised magnet at specific locations to regionally correct field distortions. The quality of shim available form passive systems depends on the range of plates and locations possible and the quality of the homogeneity correction programmes used to calculate the required plates and positioning.. Older systems could only compensate for gross field inhomogeneities but current systems can achieve very high order homogeneity. Passive shimming is perfectly stable over time with no power consumption, but it cannot compensate for short or long term changes in the external field environment, or for patient induced inhomogeneity.

     Active Resistive Shimming

    A collection of electromagnetic coils (the shim set) are fitted within the magnet bore. These coils are energised with suitable current levels. A typical shim set has 12 to 18 sets of coils which operate to distort the field in the primary axes of the magnet (X,Y,Z). For each plane there can be coils that provide a linear field distortion (1st order shim) a field that varies with the square of distance (2nd order shim) and so on, as well as inverse 1st, 2nd and 3rd order shim sets. Shimming is achieved by choosing the appropriate current for each coil MR systems using resistive shimming are subject to image degradation if there is any drift in the shim power supplies and will typically requires a stabilisation period after daily power up
    Each patient in the bore caused unique field homogeneity variations due to their body shape and particular spread of tissue and magnetic susceptibility. As the set up of a resistive shim system can be rapidly changed it can adjust for these individually induced inhomogeneities in a process known as patient shimming. Patient shimming is essential for MR spectroscopy and for achieving optimum results for frequency based fat or water suppression techniques, MTC techniques, where signal loss due to T2* relaxation times becomes significant such as high order fast spin echo, GRASE, segmented k-space, and EPI
    Patient shimming is typically automatic, and can compensate to some degree for stable changes in external environment. At least one installation has used servo linked alteration to the shim set to automatically compensate for nearby elevator traffic.

    Active Superconductive Shimming

    This approach is similar to resistive shim system, but using a shim coil set of superconductive wire fitted within the cryostat. The result is a highly stable shim with no need for constant power application. By design the superconductive shim system usually cannot be quickly adjusted to suit changing conditions.

    Gradient Off-Set Shimming

    In many systems the gradient coils are used to create the linear or first order shim corrections by carrying a small current flow at all times (offset bias). This simplifies the shim set construction and saves space in the patient bore by removing the need for a first order shim coil. The level of bias current is kept small so as not to compromise the peak performance of the gradient coils and amplifiers. Most current equipment carry out patient shimming by adjusting gradient offsets.


    GRADIENT SYSTEM

    To select the orientation of the MR scan plane, and to spatially localise the MR signals within each slice, a series of "gradient" magnetic fields are applied during the imaging sequence. These fields are called gradients because they vary linearly in space. Their effect is to strengthen or weaken the main magnetic field in a spatially predictable way.

    The application of a gradient field in conjunction with an RF pulse of controlled frequency bandwidth allows control of the location and thickness of the slice of tissues that will experience NMR (slice selective excitation). Localisation within the slice in one dimension within the excited slice is achieved by applying a gradient during echo formation and signal readout, which frequency encodes the component of the MR signal depending on spatial location (frequency encoding). Localisation in the final dimensions is achieved by analysing the effect of spatially dependant phase shifts created by transient gradient applications before signal readout (phase encoding). The Gradient system is also used to restore the phase coherence of transverse magnetisation in gradient echo and spin echo sequences.

    MR uses three gradient fields, each made to vary the field in one of the three prime axes of the machine. The fields are created by passing DC pulses (gradient current), through the gradient coils. The gradient currents are very high(200 - 600 Amps) and need to be switched on and off quickly (<< 1 ms) and accurately. This is achieved by digitally controlled high power amplifiers, referred to as the gradient amplifiers. Each of the three gradient coil sets has a dedicated gradient amplifier, the combination is controlled by a computer called the gradient control unit. Pulse shapes used to switch the gradient amplifiers are created by the host computer or by dedicated pulse control systems. Sophisticated gradient pulse control makes possible Oblique slice selection techniques to reduce motion artefacts (gradient moment nulling) and the localisation of saturation slabs.

    Due to the high current loads and heavy duty cycle, gradient systems dissipate a lot of heat. Gradient amplifiers need air cooling, some older designs require water cooling as well. Standard gradient coil sets are cooled by forced air, but most newer high powered systems use closed loop water cooling.

    The gradient coil set is secured firmly within the magnet just inside the shim coils As the current flows in the coil the interaction of the gradient field and the main field tries to distort the gradient coil or thrust it out of the bore. The flexing of the gradient coil structure is heard as the beating noise of the MRI scan. Some manufacturers entirely encase the gradient coil set in epoxy resin to quieten the system, however this approach may impedes cooling. Stronger gradients will create higher noise levels experienced in the bore, although there is some hope that better gradient coil designs will reduce noise levels. Current typical noise levels of >85 dB are already sufficient to require that patients must be given hearing protection during MR procedures.


    Radio-Frequency (RF) Systems

    Radio waves are not involved in the MRI process. The electromagnetic waves used for radio transmission and reception are a cross polarised combination of electrical and magnetic fields with the electric field component about 120 times stronger than the magnetic field. Nuclei virtually never emit radio waves, and the design of MR coils aims at produce a rotating magnetic field and little or no electrical field. For MR the "Radio" tag comes simply because the Larmor frequency of NMR nuclei fall within the section of the electromagnetic spectrum known as Radio Frequencies (10kHz-2GHz). While this discrimination may appear pedantic, it has two redeeming features. First it is a more accurate description of the process, and secondly it clarifies the description of the role of the RF system in NMR and MRI signal production by emphasising the role of a simple concept, electromagnetic induction.
    All excitation energy is provided by magnetic fields rotating in the X-Y plane, at the Larmor frequency of Hydrogen (42.58 Mhz/T). These rotating magnetic fields are created by passing RF alternating current, generated by the RF transmitter, through suitably shaped and orientated conductors known as the RF coils. This energy realigns the nuclear magnetic fields of the subject into the transverse plane and bring the nuclei fields into phase coherence by acting as a fixed magnetic field in the rotating frame. The nuclear magnetic fields rotating coherently in the transverse plane induce RF currents (MR signal) in the receive coil which are detected and analysed in the receiver system.


    RF DESIGN ISSUES

    RF Shielding

    The RF shield is a Faraday cage, ie. a closed conductive structure, that usually encloses the entire MRI examination room. This structure excludes all RF fields from the environment of the patient and the receiver coils. This is necessary as the magnitude of the received MR signal (10-9 watts) is 1000 times weaker than RF field energies typically encountered in a hospital environment. Without the RF shield these environmental fields could swamp the MR signal and imaging would be difficult. The shield can be constructed of any conductive material but non ferro-magnetic materials are preferred. Doors must have special seals fitted, windows are usually double glass laminated with a fine conductive mesh, and all cables into the room must be fitted with RF filters. Non-conductive lines can pass through tubes with an appropriate length to diameter ratio (L=5D), similar tubes are used in air-conditioning ducts and cryogen venting pipes.

    RF Transmitter Design

    The RF transmitter provides the excitation pulses for the MRI pulse sequences. It must be capable of accurately delivering pulses of RF current of controlled amplitude, frequency, and phase. Sophisticated pulse prescription components create the low frequency modulation envelopes of the required RF output. These pulses are used to modulate the output of a high quality RF power amplifier (RFPA). Typical MRI amplifiers have maximum power capabilities of between 5 and 20 Kilowatts. Solid state or valve final amplifier stages may be used with no obvious advantage in one system over the other. Amplifier control can be analogue or digital, but RF pulse shaping and timing must be digitally controlled for best results.

    RF Receiver Design

    The receive system handles the extremely low energy signals induced by the patient in the RF coils. Conventional receive systems have sufficiently high sensitivity and low noise figures for this task. The key issues for MRI receivers are stability and control of amplification, the accuracy of frequency and phase shift detection, and reference to scanner wide time base. The received signals are amplified and then demodulated in a phase sensitive detector to yield real (sine) and imaginary (cosine) signals that represent the phase and amplitude data of the MR signal.. The signals are digitised to become one line of the raw data array of an MR image (k-space). The timing of the received signal relative to the phase encoding regime of the gradient system dictates which line of Ky the data represents. Current MR systems have digitally controlled receiver systems to allow more accurate timing of operation. "Digital" receive systems vary from traditional analogue receivers by digitising the MR signal before demodulation, virtually all MR receivers have digital control of amplifiers and attenuators. Several manufacturers are equipping scanners with multiple receivers operating simultaneously to allow the use of phased array receive coils.

    RF Coils

    The coils produce a rotating magnetic field from the transmit (Tx) current and have RF current induced in them by the rotating nuclear magnetic fields of the patient during the receive phase. Coils are designed to create a magnetic field (B1) within a specific region, by the principle of reciprocity this is also the area within which time varying magnetic fields will be able to induce current flow in the coils, and the region is known as the sensitive volume. Coils design dictates that they respond preferentially to a range of frequencies, thus coils manufactured for a 0.5T machine will not work in a 1.5T scanner. Designs which respond more sensitively to a very narrow range of frequencies are known as high Q (quality) coils while lower Q coils will respond less sensitively to a broader frequency range. RF designers generally choose to pursue a high Q or low Q approach and this affects the need for accurate patient tuning of the RF coils (see coil matching below). Overall the choice of a high Q or low Q design has little discernible affect of image quality of the scanner.

    Coils and Signal to Noise Ratio

    While coils are to a degree frequency discriminating, they do not discriminate between currents induced by the transverse magnetisation representing tissue NMR behaviour (signal) and the portion of black body radiation generated by the motion of molecules (thermal noise) that happens to be aligned in the transverse plane and at the MR frequency. This noise degrades image quality in all situations. The slices selective approach to MRI causes signal to be generated only by part of the tissue in the sensitive volume of the coil but noise comes from all of the coil volume. Therefore it is best to select a coil with a sensitive volume that matches the target volume (useful field of view in 3 dimension) as closely as possible. As long as the target volume is included in the coil's sensitive volume, the smaller the coil the higher the signal to noise ratio. Typical MRI machines are available with body, head, and extremity coils and a range of anatomy specific surface coils, although the range and quality of coils varies between manufacturers and MR systems. The MR radiographers must choose the most appropriate coils for the examination to ensure optimum results.   In day to day operation, the selection of an appropriate sized coil is the primary means of assuring the best image quality.

    Polarisation


    CP versus LP  Transmission


    Coil Impedance Matching

    In all alternating current systems the simple electrical resistance of a circuit is complicated by a frequency dependant component known as reactance. The reactance of a circuit is derived from inductors (coils or turns of conductors) and capacitors (gaps, spaces, or components). and impedance. Reactance and resistance combined together represent the impedance of the circuit which is its resistance to current flow alternating at a given frequency. To allow for loss free transfer of current from the transmitter to the coil, or more crucially the coil to the receiver the impedances of these components must be matched. This is usually done by adjusting capacitors in the coil tuning circuit, which may be incorporated into each coil, or built into the scanner where coils are plugged in. systems as different patients will have varying natural Larmor frequencies which may need.


    RF CIOLS

    Transmitter Coil

    The transmit coil in particular must have a uniform sensitive field which covers the entire field of view as the strength of the B1 field must be equal throughout the volume to ensure all nuclei magnetic vectors are flipped the same degree. Many systems use all of the uniform volume coils as transmit coils, but some equipment uses the body coil as the transmit coil in all applications and optimises receive coil sensitivity at the expense of minor volume uniformity variations.

    Uniform Volume Coils

    In many applications (head, abdomen, limb joints) whole axial sections of the body need to be imaged, so a coil that delivers uniform sensitivity over its volume is desirable. The size of the coil is chosen to achieve the highest filing factor. This is usually achieved with cylindrical coil designs placed around the patient.
    The body coil is the largest uniform volume coil, fitted within the magnet bore immediately under the bore covers. Smaller diameter coils are used for extremity work and head imaging. Coil sensitivity is related to coil volume. A 30 cm. head coil will have nearly 4 times the sensitivity of the 50 cm. diameter body coil, while the smaller extremity coil can have 6 times the sensitivity. The smallest coil that can accommodate the body part to be imaged will usually give the best result.

    Surface Coils

    In many applications the transmit coil is the same as the receive coil (head, body, knee, ankle) but for a range of common examinations where the anatomy of interest is near the surface (spine, TMJ, shoulder, orbit) specialised receiver coils called surface coils provide large advantages. A receive coil can have a spatially varying field sensitivity. Most surface coils are flat and receive signals more sensitively from the regions close to them, so surface coil images appear brighter near the coil plane. This is used to advantage by excluding unwanted tissues while maximising the signal from target structures. Within its useful region a flat surface coil will exhibit sensitivity 1 to 5 times higher than a cylindrical coil that might cover the region. This increased sensitivity can be used to increase image quality or traded for increased resolution.

    Phased Array

    Phased array receive systems attempt to apply the SNR advantages of small field of view coils to relatively large fields of view by aiming a number of small receiver coils at the area of interest and combining their signals in a manner which cancels noise. They offer increases in SNR of 40% to 200% in a range of clinical applications, and should radically alter the imaging approach used in spine, neck, pelvis, breast, paediatric body, thorax and upper abdomen examinations.
    In conventional MRI systems the receiver is typically connected to one coil that is of sufficient size to cover the target field of view. With a phased array system there are multiple (typically 4) receivers. Each receiver is connected to one of a number of a small volume receiver coils constructed in an array to surround a specific body part. The coils must be designed and positioned so as not to constructively interfere in any area of their sensitive volumes, so that the noise each coil receives is not correlated to noise received by other elements of the array. Coils in this configuration are said to be decoupled. The MR signals from each receiver are combined in a way that adds the MR signal from the patient but allows the random unrelated noise components to mutually cancel. The result is the coverage of a large area with the performance of the small coils, and significant noise reduction.

    G.E. researchers P. Roemer, W. Edelstein et al first described this concept of simultaneous multiple reception as a method to extend the performance of conventional surface coils. The same workers later coined the phrase "phased array" when describing the flat style of array, and detailed techniques for collecting and combining the received signals of their four independent receive systems. Shortly thereafter a related group described a volume phased array. These three papers effectively defined the configurations of the first wave of commercially available phased array systems released by G.E. in late 1991.
    The initial phased array systems were linearly polarised (LP). Subsequently circular polarised (CP or Quad) PA has been developed. Quad PA is technically more difficult to implement than LP but can offers a further 20 to 40 percent increase over a similar LP array depending on the dimensions of the coil elements employed. Because of its high hardware component, (new receiver chains significant extra raw data memory, new coils, and extra demands on array processing) subsequently upgrading to phased array is expensive (approx $100,000 to $150,000 plus coils), but it should be clear that phased array capability is essential to any MR user seeking high quality imaging.
    More recent work has aimed at switching the received signal of 4 de-coupled coils into a single receiver in a shared time concept, but this approach seems to place severe restrictions on sequence options.

    Phased Array Coil Configurations
    Spine Array

    The spine arrays are developed from the initial flat array described by Roemer et al in 1988 and incorporating the switched array methods described by Siemens researchers at the same SMRM meeting. The individual coil elements of a switched array can be selected or de-selected via software which can even be incorporated into the scan sequence. The arrays are typically 60 cm. long and contain 6 to 9 small surface coil elements of which up to four are switched on at a time to yield two of three 40 cm. fields. To obtain good resolution, a rectangular 512 matrix is often used.
    These coils are designed to allow full spine imaging without patient handling. They are especially well suited to multiple region exams for trauma, suspected spinal block by metastatic disease, identification of spinal multiple sclerosis plaques, and the vaguely localised symptoms of spinal cord ischaemia.

    Body Array (Liver Array, Pelvic Array)

    An array of 4 surface coils, two fitted posteriorly and two anteriorly around the body for fields of view about 200 mm to 400 mm and offering at least 200% SNR increase over the conventional 60 cm. body coil. There is some inhomogeneity across the imaged volume with these variable spacing arrays, but in practice this should not be a problem. Body arrays can be used to image the liver, pelvis, thorax or as a paediatric whole abdomen coil. This type of coil should largely replace the body coil as a receiver device and may see the end of endorectal and vaginal coil usage.

    Head/Neck Array

    Designed for head and neck examinations particularly MRA. Coverage extends from the aortic arch to the top of the head and performance is significantly better that the conventional head neck array which suffers low SNR due to its high coil volume.

    Endocavity Array

    A body array incorporating an disposable endocavity coil as one of its elements to allow high resolution at the prostate or cervix with extra detail in the surrounding field of view. This product (only offered by one manufacturer) may provide a more workable image than a conventional endocavity exam, but the question will be wether adding the endocavity coil to the array supplies sufficient additional useful data to justify the intrusion.


    IMAGE PROCESSOR (Array Processor)

    The array processor converts the raw data into image data using a mathematical algorithm known as Fourier transformation. It is a dedicated computer system which communicates with the host computer drawing raw data from a temporary storage device  or RAM, and transferring the completed images to the main archiving device. The time taken to process an MRI image greatly affects the speed of the MR system in practical applications. Typical image processors take 0.05 to 0.1 second to process a 256 x 256 matrix image. As a modern scanner operating at full capacity may produce between 1000 and 2000 images per day even small differences in image processor speeds become significant.  This aspect of performance has become more significant with the increasing use of high 5122 imaging matrix (which takes 4 times as long to process as a 2562 image), and phased array images which takes 4 to 5 times as long.


    OPERATING CONSOLES

    The operators console allows the radiographer total control over scanning, image display, archive, filming, photography and patient monitoring functions. Many systems are also fitted with ancillary console for the use of radiologists, a second radiographer filming, or other clinicians. These second consoles cannot control the scanning functions and operate without reducing the speed of the main system functions. This lack of interference depends largely on the design of software and the computer network. Second consoles are commonly used for prescribing and viewing reconstructions and angiograms, and should be connected to the main image output devices.

    WORK-STATIONS

    Increasingly MRI systems are employing stand-alone image work-stations. These devices have an independent computer system but communicate with the host computer and image storage devices. They allow complex image manipulating software to be used virtually without interference to the main system. Work-stations are typically used to generate MR projection angiograms, and display these dynamically as rotating 3D perspectives, generate full volume models of the region imaged for interactive cut-aways, and arbitrary plane reconstructions form 3D data sets.


    HOST COMPUTER

    The host computer co-ordinates all functions of the MRI system. It handles all instructions from the MR operator via the MR operating software, all image manipulation and display functions, image and file managements tasks, and the creation of control pulses throughout the scanning period. Fault finding, service, quality control and external communication software is also included on many current systems. The capacity to handle several processing tasks simultaneously is determined by the architecture and capacity of the host computer in conjunction with its operating system software. Simultaneous or parallel tasking can greatly increase the speed and useability of the scanner.
    Currently available MRI scanners use host computers with a wide range of capability and speed. Recently manufacturers are acknowledging that image handling systems require large computing and storage capacity and improving this aspect of their design. It is essential to choose a machine with some redundant computer capacity and an active path for upgrades.


    PATIENT COMMUNICATION & MONITORING

    During the examination the patient is isolated in the bore of the MRI magnet for 45 minutes to 2 hours. Claustrophobia and boredom will affect many patients, so some amusement and good communication with the Radiographer throughout the examination is essential Virtually all systems have a hand held device (usually pneumatic) for allowing the patient to attract the radiographer's attention. Most systems are supplied with a room audio system carrying a voice channel and music from radio (FM), CD or tape players. The audio should also be directed to a variety of sound reducing headsets compatible with the range of RF coils used. If any of these facilities are not available on a unit, they should be retro-fitted. A video camera is often used so that the radiographer can see the patient in the bore. Devices also exist to display video (Closed circuit, TV programmes, diagnostic images) to the patient and other personnel in the room.

    PHYSIOLOGICAL MONITORS

    Most MR scanners are equipped with MR compatible ECG sets, a pneumatic bellows fitted around the abdomen or a nasal thermistor for providing a trace of respiratory motion, and a finger probe for displaying peripheral blood pulsation. These devices are primarily intended for triggering or gating scan sequences to reduce artefacts from physiological motion, or to better image that motion. The ECG and pulse probes particularly can be used in a secondary role for patient condition monitoring, but most manufacturers warn against this.    MRI compatible patient monitors are available although there is no standard code on what constitutes an MRI compatable device.   The monitors should not interfere with the MRI signal, and must not present any hazard to the patient or staff.   Care and dilligence are needed to asses the suitability of  any monitoring devices that have conductive components attached to the patient as burns have been caused at many sites.

    Issues of patient monitoring are further discussed in an associated paper on this website.


    IMAGE OUTPUT DEVICES

    MR images are purely digital, after production they exist as data stored on computer disc. This data is frequently archived to tape or optical disc. Virtually all diagnosis and clinical use is based on these images presented on large format transparent film (hard copy). As the image data contains a greater dynamic range than any visual system can reproduce the images are photographed by the Radiographer to best display the relevant information. Photography can make or break an examination, and is a basic skill of every radiographer working with a digital modality. Hard copy is produced on a laser camera, or multi-format video camera.
     


    REFERENCES

    Thomas S.R. "Magnets and Gradient Coils: Types and Characteristics." 1992 AAPM Summer School notes P.54
    GE PROFILE and Hitachi STRATIS Trade display release RSNA 1994 Chicago
    MRT-22A sales brochure Toshiba Medical Imaging June 1987
    Simon H.E.,Carson A.N. Dynamic Effects of Moving Ferrous Objects on MR Image Quality. Abstracts of SMRM Meeting 1988 P. 1062
    Andrew E.R. & Szczesniak E. Magnetic Shielding of Magnetic Resonance Systems MAGNETIC RESONANCE IN MEDICINE Vol 10 Pp.373-387 1989
    Hudson F.R., Colvin J. Et al High Field Superferric NMR Magnet Abstracts of SMRM Meeting 1988 P. 23
    Andrew E.R. Passive Magnetic Screening. Magnetic Resonance in Medicine Vol. 17 Pp. 22-26 1991
    Sobol W. General Site Requirements including RF & Magnetic Shielding. The Physics of Magnetic resonance Imaging 1992 AAPM Summer School.
    Mansfield P., Chapman.B., et al Active Acoustic Screening: Reduction of Noise in Gradient Coils by Lorentz Force Balancing. Magnetic Resonance in Medicine. V33 No. 2 P. 276-281 (1995)
    Walling J. Noise and the Technologist. SIGNALS No 5 Feb 1993 P.6
    Hoult D. The Radio Wave Myth. MR PULSE V1 No.3 P 27 Dec.1994
    Roemer P.B., Edelstein W.A., Souza S.P., Hayes C.E. Mueller O.M. "Simultaneous Multiple Surface Coil NMR Imaging" Abstracts of SMRM Meeting 1988 P. 875
    Roemer P.B., Edelstein W.A., Souza S.P., Hayes C.E. Mueller O.M." The NMR Phased Array" MAGNETIC RESONANCE IN MEDICINE V16 Pp 192-225 1990
    Hayes C.E., Hattes N. Roemer P.B. "Volume Imaging with MR Phased Arrays" MAGNETIC RESONANCE IN MEDICINE V18 Pp. 309-319 1991
    G.E. Product information brochure 1991
    Requardt H. Erhard J., Offerman W., Loeffler W. "Fast switched array coils: Applications to Multislice/Multicoil Sequences
    RESONANCE TECHNOLOGY Co. 22647 Ventura Boulevard, Suite 222, Woodlands Hills California 91364 USA