29

*Corresponding author, Phone: ＋1-216-368-3894, Fax:＋1-216-368-4671, E-mail: hiroyuki.fujita＠case.edu

29

Magn Reson Med Sci, Vol. 6, No. 1, pp. 29–42, 2007

REVIEW

New Horizons in MR Technology: RF Coil Designs and Trends

Hiroyuki FUJITA1,2,3*

Departments of 1Physics, Case Western Reserve University, and 2Radiology, UniversityHospitals of Cleveland, 10900 Euclid Avenue, Cleveland, Ohio 44106-7079, USA

3School of Information Technology and Electrical Engineering,The University of Queensland, Brisbane Qld 4072, Australia

(Received January 9, 2007; Accepted February 14, 2007)

Parallel imaging techniques have developed very rapidly, and realization of their fullpotential has required the design of magnetic resonance (MR) scanners with ever-increasingnumbers of receiver channels (32 to 128). In particular, 1.5- and 3-Tesla fast MR imagingapplications are now used in everyday clinical practice. Both strengths require maximumachievable signal-to-noise ratio (SNR) and multi-detector array coil optimization withinthe framework of the parallel imaging scheme for more advanced and faster clinical MRscanning. Preampliˆers are key components in the detector array coils and serve manyfunctions beyond mere signal ampliˆcation. One critical function is to aid in thedecoupling of individual coils, which is essential for optimal SNR and the performance ofparallel imaging. To support a large number of detector array coils for parallel imaging,preampliˆers must be physically very small so that they may be tightly packed together toform an optimized detector array. The author herein reviews the state-of-the-art workreported by those skilled in the art to consider the rationale for determining how manychannels are enough and how fast we can go. The paper explores the important andfundamental principles of RF array coils for MR imaging and reviews cutting-edge arraycoils, including those for transmit-SENSE or parallel transmission applications. The futureof radiofrequency (RF) coil technology is also considered.

Keywords: phased array RF coils, preampliˆer decoupling, parallel imaging, paralleltransmission, wireless RF coils

Introduction

Modern radiofrequency (RF) array coil design iscomplex, and its understanding requires broadknowledge of magnetic resonance (MR) physicsand electronics. Today's coils are very intricate anddelicate. For example, as many as 400 to 1500individual componentsWparts are used in the con-struction of 8-channel head coils and 16-channelbrain-spine combo coils, both of which are typicaldetector array coils. Every component must becorrectly integrated; if even one component isbroken or damaged, the coil does not functionoptimally. The author attempts to explain thecritical components that constitute an RF arraycoil, especially emphasizing principles and methodsfor decoupling array coils, which is essential to

today's advanced and sophisticated clinical appli-cations.

In 1978, in his historic paper, Hoult detailed anddiscussed the nuclear magnetic resonance (NMR)receiver;1 in 1990, Roemer and associates proposedthe phased-array coil technology that underlies thedesign of most of today's RF receive coils;2 and in1991, Hayes's group discussed the application ofthat technology to volume imaging.3 Original andgeneral phased-array radar technology4,5 and arraycoil theory and application6 are reviewed in detailelsewhere.

A small surface coil is well known to yield higherSNR nearer the surfaces of the coil than does avolume coil, such as the birdcage coil,7 but theB1-sensitive region of a small surface coil is (much)smaller than that of a volume coil. It is remarkablethat in an approach utilizing phased-array coils, thehigh signal-to-noise ratio (SNR) associated with asmall surface coil can be achieved and maintained

30

Fig. 1. The mechanism of electromagnetic coupling between the patient andthe magnetic resonance (MR) radiofrequency (RF) coils, due to the straycapacitances present, is illustrated. Baluns are added to block the commonmode current and allow the diŠerential mode current to ‰ow.

30 H. Fujita

Magnetic Resonance in Medical Sciences

over an extended ˆeld of view (FOV) by using a setof RF coils. This is possible because of the develop-ment of various array coil designs for imagingapplications of interest and of techniques to decou-ple the mutual inductances and thereby elimi-nate cross-talk among the elements of the arraycoils. Adjacent coil elements are often partiallyoverlapped to cancel the mutual inductance be-tween the elements. The next nearest and othercoil elements can be mutually decoupled using alow-input impedance (typically less than 5 Ohms)preampliˆer decoupling circuit. Other decouplingtechniques and methods are brie‰y discussed later;detailed discussions of low noise ampliˆers may befound elsewhere.8

In conjunction with a coil matchingWdecouplingcircuit, the low-input impedance preampliˆereŠectively eliminates current ‰ow in the coil ele-ment loop, thereby eliminating the magnetic ˆeldsinduced in neighboring coil elements (shown later).Baluns and considerations pertaining to cablerouting are also important factors in array coildesign. Figure 1 illustrates how cables couple withthe patient and why baluns need to be implement-ed; details are discussed elsewhere.9

RF coil technologies are driven by the rapidadvances in development of MR scanners withgreater numbers of receiver channels as well as indevelopment of parallel imaging techniques;10–12 the96-channel head array coil is one such example.13

Whatever the application at any ˆeld strength,maximum achievable SNR is key. Because SNR isdestined to be degraded or lowered from thestarting point, i.e., the signal source, to the endpoint of the receiver chain, minimizing its loss inthe process is of primary concern. In this paper, thecomponents of an array coil are explained, andissues requiring special attention are noted. The

author uses state-of-the-art examples to address thequestions of what constitutes a su‹cient number ofchannels and how coil elements in an array aredecoupled and to consider how array detectors canbe optimized.

Signal-to-Noise Ratio

Signal-to-noise ratio (SNR), one of the mostimportant parameters to be optimized in MRimaging applications, was detailed by Hoult andRichards.14 Here, it is essential to understand howSNR relates to the coil-related parameters. Thesignal is written as

S1g3B 20B 1

xy(r ) Eq. (1),

where g is the nuclear gyromagnetic ratio, B0 is thestatic magnetic ˆeld, and B1

xy(r) is the RF magneticˆeld produced by a coil with unit current 1A. Indesigning a coil, B1

xy(r) must be optimized, whichrequires the appropriate choices of coil size over atarget FOV and of distance from the FOV. Sizechoice depends generally on the number of availa-ble receiver channels.

On the other hand, NMRWMR imaging noise isthermal noise, and the noise generated from thecoil is given by

N＝ 4kTDfR Eq. (2),

where k is the Boltzmann constant, T is the temper-ature in K, Df is the bandwidth, and R＝RC＋RS.RC is the coil resistance and RS the sample energyloss, i.e., the equivalent series resistance resultingfrom the induced eddy current losses in the conduc-tive sample. Combining Eqs. (1) and (2), SNR canbe expressed using the coil-related parameters as

31

Table 1. Relationship between dB expression and common factor

Factor 1W100 1W10 1W5 1W3 1W2 2 3 4 5 6 7 10 100

dB －20 －10 －7 －4.8 －3 3 4.8 6 7 7.8 8.45 10 20

Fig. 2. Depiction of a preampliˆer noise model. Vdenotes the signal; NS, the noise generated by theresistance, r, of the input signal source; and NP, thenoise generated inside the preampliˆer.

31RF Coil Designs and Trends

Vol. 6 No. 1, 2007

SN

1B 1

xy(r )RC＋RS

Eq. (3).

To optimize SNR, B1xy(r) can be maximized by

having the coil closer to the sample, and RS can beminimized by matching coil size to the target FOV.In fact, Wright and Wald compared an N-elementarray to a single coil with respect to the resultantSNR while overall coil dimension remained un-changed.6 They showed, for example, that whencomparing a single coil with arrays measuring 8×8,4×4, and 2×2, improvement in SNR disappearsat a depth equal to the diameter of the array.However, for distances in between, SNR curvesvary according to the number of array (i.e., coil)elements. If each coil is made too small, the coilresistance loss dominates the sample noise. Thisimplies that optimization is required for a givenspeciˆc application, e.g., target depth and availablenumber of receiver channels. RC is minimized bymaking the unloaded coil Q high. It is well knownthat at low-B0 ˆeld systems (Ã0.3T), coil resistanceis dominant or at least comparable to sample noise,but at 1.5T and above, sample noise dominates(i.e., RS:RC). Furthermore, it is useful to be awarethat if each of 2 coils yields the same relativesensitivity in free space and is dominated by samplenoise, the two have the same absolute sensitivity.15

(The detailed discussion of the coil unloaded Q,loaded Q, and sensitivity in this reference should beuseful to the coil engineer.)

Low Noise Preampliˆer

The characteristic parameters of preampliˆers,such as the noise ˆgure (NF), aŠect SNR perfor-mance. Thus, a preampliˆer is a key component inan RF coil and plays a critical role in the design of adetector array coil. A brief summary of the pream-pliˆer's function follows. The electromotive forceor induced voltage (i.e., signal) in a coil is verysmall, typically on the order of a few mV. A pream-pliˆer increases this small signal to a few mV, e.g.,30 dB (i.e., 1000 times) greater. In practice, gainusually refers to power gain; thus, a 20-dB gain inpower corresponds to a 10-dB gain in voltage.(Quick conversion factors are shown in Table 1.)

The NF, one parameter for measuring the per-formance of the preampliˆer, is detailed else-

where.1,16 Figure 2 depicts a simple model forpreampliˆer noise. V denotes the signal; NS, thenoise generated by the resistance, r, of the inputsignal source; and NP, the noise generated insidethe preampliˆer. Using these quantities, NF isdeˆned as

NF＝10 log10 ØN 2P＋N 2

S

N 2S » where NS＝ 4kTDfr

Eq. (4).

The industry standard for preampliˆer NF is lessthan 0.5 dB, and the ˆrst NF and gain are wellknown to have the most signiˆcant impact in theentire electronics circuit, which is often cascaded.Kucera, Lott, Horowitz and Hill demonstrate thisin their respective chapters on the design of low-noise ampliˆers.8,16 Because noise is generated inany passive elements that dissipate power, includ-ing cables, the design of most detector array coilsintegrates preampliˆers so as to minimize undesira-ble noise. The multi-detector array is a key compo-nent for parallel imaging and comprises many (32to 128) detectors. As stated, preampliˆers not onlyamplify signals; they are key features of thesedetectors that assist in the decoupling of individualdetectors, and decoupling is critical to optimizingparallel imaging. The very small size of the pream-pliˆers allows them to be tightly packed to form anoptimized array.

A Single Coil Model

To understand the mechanism underlying thefunction of detector array coils, we start with aschematic representation of a single coil thatincludes a simpliˆed preampliˆer decoupling circuit(Fig. 3[a]). This coil circuit is a building block of anarray coil. L1 represents the coil inductance and

32

Fig. 3. (a) A single-coil model with a low-input im-pedance preampliˆer. (b) A system of 2 coils coupledthrough non-zero mutual inductance. (c) A system of 2coils coupled through non-zero mutual inductance. Alow-input impedance preampliˆer is added to Coil 2 forpreampliˆer decoupling. (d) The matching inductor L2

(L2A＋L2B) plays several critical functions in preampliˆ-er decoupling. Each quarter-l circuit has its characteris-tic impedance for necessary impedance matching.

32 H. Fujita

Magnetic Resonance in Medical Sciences

R1, the coil resistance (typically around 0.5 Ohms inair and 5 Ohms when placed on a phantom). C1

and C2 are the respective tuning and matchingcapacitances. When looked at from the side of thepreampliˆer (i.e., that side of the circuit), a part ofL2, a matching inductor, functions to match thecoil impedance to 50 Ohms together with C2. At thesame time, L2 achieves a parallel resonant circuitformed with C2, in particular, when the inputimpedance of the preampliˆer, rpreamp, becomessmall (e.g., 0 to 2 Ohms). When looked at from theside of the coil, impedance is inˆnity, so the circuithas high impedance and thus is equivalently open.It is extremely important to note that becausecurrent cannot ‰ow in an open coil, no magneticˆeld can be induced through non-zero mutualinductance in neighboring coil elements. Thisprinciple underlies the elimination of cross-talkamong coil elements. This decoupling mechanism isexplained in more detail later, but it is worthmentioning that the original signal source changesfrom the current source (without preampliˆerdecoupling) to the voltage source (with preampliˆerdecoupling) that contains the necessary informa-tion without losing the integrity of the originalsignal information.

Preampliˆer Decoupling

We now consider a system of 2 coupled coils andthe underlying mechanism of their decoupling. InFig. 3(b), Coils 1 and 2 are coupled through mutualinductance, M12. Coil 1(2) is represented by its self-inductance, L1(2); resistance, R1(2); and capacitanceC1(2). Vsignal indicates the signal induced in Coil 1,and Vout indicates the output voltage. When the ACcurrent, I1, ‰ows in Coil 1, the induced AC current,I2, will ‰ow in Coil 2 through the non-zero mutualcoupling between the 2 coils. Thus, assumingharmonic time dependence (i.e., e－ivt), the outputvoltage is expressed as

Vout＝Vsignal＋ØR1＋iØvL1－1

vC1»»I1＋ivM12I2

Eq. (5).

This equation is essential to understand the needfor the coil to be resonated at the target Larmorfrequency and to comprehend the mechanism ofpreampliˆer decoupling. The second term in theright side in Eq. (5) associated with I1 may berealized as the noise related to Coil 1, and the thirdterm associated with I2 is recognized as the noiseresulting from the coupling between Coils 1 and 2.The noise related to Coil 1 can be minimized bytuning and matching Coil 1 for resonance, i.e.,

33

Fig. 4. (a) Schematic representation of a detectorarray coil (8-channel). (b) Depiction of B1 sensitivityproˆle in an 8-channel array coil in arbitrary units.

Fig. 5. The structure of a micro-strip line coil. Toachieve necessary impedance matching, the thick-ness, h, of the dielectric medium and the width, w, ofthe conducting strip are adjusted such that thecharacteristic impedance Z0＝50 V.

33RF Coil Designs and Trends

Vol. 6 No. 1, 2007

the imaginary part of the term associated with I1

vanishes, leaving the intrinsic real resistance, R1.Decoupling requires that the third term in the rightside in Eq. (5), ivM12I2, i.e., the noise through themutual coupling, be zero. To satisfy this condition,we can arrive at 2 cases: either M12＝0 or I2＝0.When M12＝0, Coils 1 and 2 are overlapped tonullify the mutual inductance between the two.Capacitive decoupling may be employed here tocancel the mutual inductance.17 The principlebehind the capacitive decoupling is that the non-zero mutual impedance is modeled as an induc-tance, and the inductance is canceled by the addi-tion of capacitance (i.e., the 2 coils are connectedby a capacitor). I2＝0 corresponds to the case ofso-called preampliˆer decoupling.

Now Coil 2 is integrated with a preampliˆerdecoupling and matching circuit (Fig. 3[c]). Here,L2 is a matching inductor, and rpreamp represents theinput impedance of the preampliˆer; as rpreamp

becomes 0, the inductor, L2, and the capacitor, C2,can be chosen to form a parallel resonant circuitat the target MR frequency, yielding a highimpedance (i.e., inˆnity, theoretically speaking).Coil 2 then becomes an open circuit, and there is nocurrent ‰ow, i.e., I2＝0. At this stage, even if themutual coupling between the 2 coils is not zero,Coils 1 and 2 are decoupled using a low-inputimpedance preampliˆer. This is the art of preampli-ˆer decoupling. Furthermore, it is emphasized thatthe matching inductor, L2, denoted as L2A and L2B

in Fig. 3(d), plays several crucial functions thatmay not appear obvious. L2A functions to matchthe coil impedance to 50 Ohms together with CM

when looked at from the side of the preampliˆer(IIIªII). At the same time, L2A (and L2B and C2,which are adjusted to be ``short,'' i.e., the seriesresonance) achieves a parallel resonant circuitformed with CM, in particular, when the inputimpedance of the preampliˆer, rpreamp (approximate-ly RL2B＋RC2), becomes small (e.g., 0 to 2 Ohms).When looked at from the side of the coil (IªII),the impedance is inˆnity, thereby having a high-impedance or open circuit. In other words, thedecoupling circuit considered here and shown inFig. 3(d) (sandwiched by 2 dashed lines) is equiva-lently a mismatched lW4 (quarter wavelength cable)circuit; one end is open and the other end is short.It is instructive and meaningful to view the circuit(regions denoted as II and III) such that each regionmay be represented as a corresponding lW4 circuitwith its characteristic impedance, Z0. A transmis-sion line that is an odd multiple of quarterwavelengths long that terminates in an impedanceZLoad presents an input impedance of

Zin＝Z 0

2

ZLoadEq. (6),

where Z0＝characteristic impedance. In turn, Z0＝ZinZLoad. In other words, by appropriately choos-

ing the characteristic impedance of the matchingsection, a quarter-wave section can be used tomatch any 2 impedances, such as in sections II andIII of Fig. 3(d). The transformations in impedancefrom I to II to III and in reverse are executedcorrectly and consistently because each region has

3434 H. Fujita

Magnetic Resonance in Medical Sciences

its associated and appropriate characteristic im-pedance. With this quarter-wave matching sectionmechanism, we turn to the preampliˆer.

A preampliˆer is noise matched; for a given ˆeld-eŠect transistor (FET), there is a signal-sourceimpedance that optimizes or minimizes the noiseˆgure. The optimum NF can be achieved when theimpedance of the receive coil matches the sourceimpedance of the FET. Because the FET im-pedance is relatively high (:50 Ohms), animpedance transformer comprising L2B and C2

together forms a lW4 circuit in section III ofFig. 3(d) and transforms 50 Ohms to the high inputimpedance of the FET for the lowest NF. Anoptimal impedance for the lowest noise ˆgure oftypical FET is on the order of a few hundred V.Thus, L2B and C2 transform the coil's 50 Ohms tothe high impedance of FET for the lowest NF. Asan additional note, the input impedance of thepreampliˆer in Fig. 3(d) is approximated as the sumof resistances associated with L2B and C2. However,since the Q of the capacitor is typically much higherthan that of the inductor, the input impedance isdominated by the contribution of the resistanceassociated with the inductor.

With an understanding of this preampliˆerdecoupling method, it is straightforward to extendthe array concept to 8-, 16-, and even larger num-bers of channels. Figure 4(a) shows an 8-channelarray coil in which each coil has its own B1 sensitivi-ty proˆle (Fig. 4[b]). Because it is a surface coil,which yields a high SNR over only a small region,an overall high SNR over the target FOV or regionof interest (ROI; Fig. 4[b]) can be generated if eachchannel SNR is combined by a sum-of-squaresmethod,2 for example. As discussed, taking intoconsideration the appropriate size of each coil withSNR optimization, the number of coil channels isdetermined by the number of available receiverchannels and the target FOV or ROI. Many refer-ences discuss diŠerent array coil designs for variousimaging applications at diŠerent ˆeld strengths.Although a detector array coil using quadraturepairs (for each channel) is possible, as the numberof available receiver channels increases, the currenttrend is to use a set of various (size and shape) loopcoils.

Other Array Coil Designs

This paper focuses on detector array coil designsthat employ preampliˆer decoupling because theseare the most popular and widely accepted designs at1.5 and 3 Tesla. Other decoupling techniquesinclude: overlap;2 quadrature (intrinsic isolationW

decoupling); solenoidal array employing anti-turn loop;18 capacitor decoupling to cancel induc-tance;17,19 decoupling network after coils to mani-pulate the electric ˆeld induced by the mutual in-ductive coupling;20 and shield or transmission linedesigns, which are often employed for very highˆeld applications (Fig. 5) because most electromag-netic ˆelds are contained between the coil elementand the shield.21,22 In these shielded or transmissionline designs, for imaging, electromagnetic-ˆeldleakage is small, and because the leaked ˆelds areonly a small portion of all ˆelds, the isolationbetween coils remains adequate even if leakedelectromagnetic ˆelds couple with each other.

Various Parallel Imaging Methods (SMASHversus SENSE)

With the advent of parallel imaging tech-niques,10–12,23 array coil design plays a major rolein achieving desired parallel-imaging perfor-mance.24–33 However, with respect to coil design,there is a major diŠerence between simultaneousacquisition of spatial harmonics (SMASH), a``k-space'' reconstruction approach, and sensitivityencoding (SENSE), an ``x-space'' (i.e., imagedomain) reconstruction approach. In SMASH,spatial harmonics are achieved by manipulating theB1 sensitivity of each coil in the array to replacesome k-space data lines that would otherwise havebeen collected by the application of conventionalgradient phase encoding. This requires that thedirections of the array and desired phase encodingcoincide. In clinical practice, the complex shapesand curvature of the body often make it di‹cult toconstruct a SMASH array coil whose directionWorientation coincides with that of the desired phaseencoding, at the same time generating the spatialharmonics required in that direction. In contrast,SENSE is less restrictive about placement of eachchannel coil in conjunction with a chosen phaseencoding direction. However, SENSE requires thatone be able to tell which B1 sensitivity belongs towhich coil so that the aliased image that is createdunfolds and the desired acceleration factor can beachieved. The acceleration factor is the inverse ofthe reduction factor in imaging time. The measureof ease to distinguish B1 sensitivity is given by aso-called g-factor, and for a given imagingapplication, coil design requires that the g-factor beoptimized.34 Furthermore, the acceleration factordepends upon the number of receiver channels;e.g., to achieve an acceleration factor of 3 in x-direction, there must be at least 3 individual coils inthat direction. Additionally, Ohliger and associates

35

Fig. 6. (a) Prototype coil of 3T, 128-channel cardiac array coil. (CourtesyLawrence L. Wald, Massachusetts General Hospital). (b) A 7T base shieldedbirdcage TWR brain coil measuring 18 cm long with a diameter of 27 cm. Theshield protects against radiation loss at very high frequencies. (Courtesy John L.Patrick, Philips Medical Systems). (c) A 7T, 24-channel whole-brain array.Twenty-four elements are arranged in 8 radially gapped columns (30z gapbetween elements). The array has anterior columns (2 elements) and posteriorcolumns (4 elements). The image is obtained with MPRAGE: repetition timeWecho time (TRWTE) 5W11 ms; 512×512; 2-mm slice; 0.5×0.5×2 mm3 voxels.(Courtesy Patrick J. Ledden, Nova Medical, and JeŠ H. Duyn, NationalInstitutes of Health [NIH], Bethesda, MD). (d) A 7T, 16-channel volume striparray with 16-way TWR interface. (Courtesy Ray F. Lee, New York University).

35RF Coil Designs and Trends

Vol. 6 No. 1, 2007

attempted to establish an upper bound on thespatial encoding capabilities of array coils inparallel imaging.35 A few highlights from theirwork include that: a limit for acceleration is 5× for1D undersampling of k-space; a net accelerationfactor of 15× is the limit for 2D; a higher accelera-

tion factor can be achieved as B0 becomes higher(due to shorter wavelength); and tissue conductivityaŠects parallel imaging performance. The achieva-ble SNR with ˆnite coil arrays is discussed in.36

These insightful considerations should be helpful tocoil designersWengineers.

36

Fig. 7. Illustration of the 3T QD birdcage transmit and 8-channel receive hybrid coil and3T knee images obtained with the coil. (left: 3D SPGR, a＝39; right: 3D SPGR, a＝189) (Im-ages Courtesy Brian Rutt, Robarts Research Institute)

Table 2. The diŠerences between parallel imaging and transmit-SENSE

Parallel Imaging Transmit-SENSE(parallel transmission of

multidimensional RF pulses)

What do we know? Aliased images,coil sensitivities

Full ˆeld-of-view excitation,coil sensitivities

What do we want to know? Full ˆeld-of-viewimages

Aliased excitations

Can we use preampliˆer decoupling? Yes Noªother decouplingtechniques needed

36 H. Fujita

Magnetic Resonance in Medical Sciences

State-of-the-Art Examples

One-and-a-half-Tesla MR imaging systems arethe gold standard in many clinical applications, andsome systems are now available with 96 RF receiverchannels.13 3T systems have proven their advan-tages in brain and musculoskeletal imaging, andrapid development in body applications are inprogress. The 3T market is being recognized as agrowing high-end clinical market, and some 3Tsystems are now equipped with 128 RF receiverchannels (Fig. 6[a]). On the other hand, a few 7Tsystems are available worldwide that are targetingthe study of brain function; functional brain imag-ing requires a ˆner spatial scale, with resolutions ata sub-millimeter scale. Researchers are hoping thatsuch brain analysis enables improved treatment ofneurodegenerative diseases, such as Alzheimer'sdisease. Because of the available intrinsic high SNRat 7T (Fig. 6[b] and [c]),37 images based on 30 ormore physicochemical parameters are used to exa-mine tissue morphology, blood ‰ow, metabolism,and chemistry in vivo. Many diŠerent RF detector

array coils are available at these diŠerent staticmagnetic-ˆeld strengths. Figure 6(d) shows anexample of shieldWtransmission-line type designs,referred to as volume-strip arrays.38

Nonetheless, the essential questions remain:what are the clinical applications; how manychannelsWcoils do we have available and use; howdo we decouple the coils; and how do we optimizethe SNR?

Another important consideration is the speciˆcabsorption rate, SAR, which is the quantity charac-terizing the energy deposited into a unit mass:39

SAR1v02B 1

2a2 Eq. (7),

where v0＝Larmor angular frequency, B1＝trans-mit RF magnetic ˆeld amplitude, and a＝coil di-ameter (or patient size). As the static magnetic ˆeldincreases from 1.5 to 3T, the resultant SAR at 3T is4 times that at 1.5T. Thus, at higher ˆeld strengths,SAR becomes the issue for particular sequencesthat require a large ‰ip angle andWor many slices tobe collected within the repetition time (TR). In suchcases, a hybrid coil can be advantageous. A hybrid

37

Fig. 8. A 9.4T, 400 MHzTWR multi-element TEMparallel transceiver headcoil for B1 control. (Court-esyThomasVaughan,Uni-versity of Minnesota)

Fig. 9. (a) Illustration of direction of ‰ow and location of aliased information. (left:parallel imaging; right: Transmit-SENSE). (b) Normal excitation versus inner volumeexcitation. (c) Example of homogeneity correction. (Courtesy Mark A. Griswold, CaseWestern Reserve University)

37RF Coil Designs and Trends

Vol. 6 No. 1, 2007

coil consists of a localized dedicated transmittercoil whose size is just optimized for a target ROIWFOV and an independent set of receiver coils asshown in Fig. 7.40 Because the transmitter coil ismuch smaller than a typical whole body transmittercoil, the localized transmitter requires less inputpower to generate the necessary B1 or a much largerB1 can be generated that translates into a shorterRF pulse duration and potentially reduced SAR.Furthermore, Eq. (7) means that an RF coil withhalf the diameter yields one fourth the SAR provid-ed that B1 remains unchanged. For this reason, thehybrid coil is considered an eŠective approach athigher static magnetic-ˆeld strengths.

38

Fig. 10. The world's ˆrst simultaneously obtained MR-PET human brain im-ages. The patient was injected with FDG, and the PET acquisition took 20 min.During the PET acquisitions MR images with T2 (a) and Dark Fluid contrast (b)were acquired. The images show the ˆndings: periventricular gliotic foci in thewhite matter mostly present in the bioccipital region, and calciˆcation-irondeposition of the basal ganglia. (Courtesy Siemens Medical Solutions, Univer-sity of Tennessee, and University of Tuebingen)

38 H. Fujita

Magnetic Resonance in Medical Sciences

Transmit-SENSE

At very high frequencies, e.g., 3T and above,wavelength eŠects within a dielectric object havebeen reported to produce non-uniformity issues, asseen in dielectric resonance in which object dimen-sion and wavelength coincide. Thus, achievingB1 transmission to yield a uniform B1 ˆeld is chal-lenging. This is why some researchers attempt tocorrect the B1 ˆeld by manipulating the magni-tudes and phases associated with each transmitterchannel, known as ``mode-scanning excitation''(Fig. 6[d])38 and ``B1 shimming'' (Fig. 8).41

However, this B1 shimming process is an iterativeprocess, and keen insight is required to control B1

spatial modulation. Transmit-SENSE,42 on theother hand, is an inverse approach. It is well knownthat a Fourier transform relationship exists in thereception (omitting constants for simplicity) be-tween signal and image (i.e., object):

M( …x)＝fS(k)e－i …xkdk Eq. (8).

Katscher's group realized that, in transmission,there is a Fourier transform relationship betweenthe excitation RF pulse shape and the 3-D excitedspatial proˆle:

Mexcite( …x)＝fB1(k)e－i …xkdk Eq. (9).

In the Transmit-SENSE approach, diŠerent time-varying B1 pulses in each coil of the array are usedto excite a multi-dimensional proˆle in a reduced

time.43

Table 2 illustrates the diŠerences betweenparallel imaging and Transmit-SENSE. Withregard to RF coil design, the major diŠerencebetween SENSE and Transmit-SENSE is thatSENSE can utilize preampliˆer decoupling amongthe detector array coil elements, but Transmit-SENSE cannot because the current is required toproduce a B1 transmit ˆeld. Thus, learning how todecouple multiple transmit coil elements stillremains a challenge. Kurpad's and Vernickle'sgroups have presented examples of coil designs forTransmit-SENSE.44,45

As recognized in Table 2, with respect to direc-tion of ‰ow and location of aliased information,process ‰ow is opposite for parallel imaging andTransmit-SENSE. The ‰ow is depicted in Fig. 9(a).Use of multi-dimensional RF pulses enables follow-ing applications such as inner volume excitation(Fig. 9[b]; faster imaging in localized regions) andhomogeneity correction (B1 shimming; Fig. 9[c]).By designing RF pulses to compensate for transmis-sion inhomogeneity and susceptibility correction,which is for B0, inhomogeneity is corrected.

EmergingWFuture RF Coil Technologies

The healthcare industry and medical technol-ogies are rapidly advancing, and each medicalspecialty may be supported by a large industrialeŠort. However, because of the variances in func-tion between specialties, it is anticipated that manydiŠerent medical technologies will be integrated to

39

Table 3. Various imaging modalities

ImagingModality

At what arewe looking?

IonizingRadiation Pros Cons

ComputedTomographyWX-ray

Electron density Yes Fast scan time,high spatialresolution,widespreadaccess

Radiationexposure,compromisedsoft tissuecontrast

PET (NuclearMedicine*):imaging withpositron emitters

Injection of aradioactivelylabeled drug thataccumulates intarget organs orpathologictissue, tracerslabeled withpositron emitters

Negligible(equivalent toamount wereceive fromnaturalenvironmenteach year)

Characterizationof biochemicalfunctions ofcells, organs, andbody structuresin vivo,metabolicactivity at amolecular level

Compromisedspatial resolution(blurring), aboutone hour to waitfor the injectionto go through thebody prior toimaging

MagneticResonanceImaging

Proton density(mostly)

No High spatialresolution,excellent softtissue contrast

Patients withpacemakers,claustrophobia,acoustic noise

SPECT(NuclearMedicine*):imaging withsingle photonemitters

Radiotracers inthe human bodywith variousdecay energies

Negligible Simpler scannerthan PET

Compromisedlocalizationcapabilitycompared withPET

Ultrasound Propagation ofsound wave(f＝1 to 15MHz) into thepatient (backscattering)

No Portability,real-timeimaging

Image qualitytoo dependent onspeciˆc patient

*Nuclear medicine involves the emission of radiation from the organ or tissue, whereas radiology involves thetransmission of radiation into the tissue.

39RF Coil Designs and Trends

Vol. 6 No. 1, 2007

realize more powerful and useful diagnostic tools.The principle underlying the combination ofimaging modalities, such as PET-CT,46 is that eachprovides diŠerent image contrasts and givesdiŠerent information. In this example, because CTreveals the shape of cancerous tissue and PETshows the activity of the cancer, PET and CTtogether simultaneously reveal information abouttumor shape, location, and activity. SiemensMedical Solutions (Erlangen, Germany) is workingon combining the strengths of MR (soft tissuecontrast) and PET (molecular information),47 asreported in Fig. 10. Table 3 summarizes the variousimaging modalities. Successful integration requiresother modalities to at least be compatible with MR,i.e., non-magnetic, and RF coil designs will have to

take the requirements of other modalities intoconsideration. For instance, Kuroda is working onan MR-compatible endoscope made of titanium,glass, and polymers with a 1.5-cm ID for a tip RFcoil.48 This yet unknown and emerging ˆeld willfurther constrain the design of MRI RF coils.

The possibility of wireless RF coils is worthmentioning. As discussed earlier and elsewhere,9

cabling is a challenge in the design of RF receivercoils with many channels because of couplingsamong the cables49 and coil elements, space limita-tions, and weight. Obviously, a cable with galvanicinterconnections exists between an RF coil and MRimaging system to: send the MRI RF signal withits high dynamic range (typically 150 dB withmodern systems) and bandwidth; control signals

4040 H. Fujita

Magnetic Resonance in Medical Sciences

for tuningWdetuning, recognizing coil ID and coilstatus; and supply power to the onboard compo-nents. Wireless technologies may advance safety byeliminating baluns and oŠering convenience andease of operation for particular clinical applica-tions (e.g., in the use of the anterior cardiac arraycoil). On the other hand, challenges include ˆndingmethods to maintain the phase information of in-dividual signals.50 This can be done in the analogdomain by mixing the RF signals on to higher fre-quency carriers, or it can be done digitally if eachcoil is equipped with its own AWD converter.Another issue is that the necessary wireless datatransmission can be greater than 1 GBWs for a high-end MR imaging scanner with high receiver-chan-nel count, assuming the required bits accuracy,sampling rate, maximum gradient strength, andFOV. This speciˆcation is not yet available fromthe current WLAN industry. Furthermore, the im-plementation of wireless technology requires a bat-tery to power the coil and its associated on-boardelectronics. However, it is anticipated that theaforementioned challenges will be overcome be-cause of the rapid technological advancement in theindustry, as evidenced by the growth of the wirelesscommunication industry. Thus, the challengeremains as to how we can take advantage of theseemerging technologies.

Summary

The subject of array coil design is complex andencompasses the entire reception chain andemerging applications of parallel imaging andtransmission technologies. Although there aremany coil designs for diŠerent applications atvarious B0 ˆeld strengths, the principles underlyingthe design of detector array coils are essentially thesame. In particular, preampliˆer decoupling inview of a mismatched lW4 matching circuit isimportant. Design of RF detector array coils musttake into consideration that: (1) SNR be optimized;(2) for a given clinical application at a selected B0,the number of available receiver channels, desirablephase encoding direction, target acceleration fac-tor, and FOV determine the coil design of choice;and (3) SNR (for spatial resolution) and speed ofacquisition (for temporal resolution, i.e., accelera-tion factor) are best balanced depending upon clini-cal application.

Additional integration of diŠerent imagingmodalities is anticipated, which will further con-strain design of RF array coils intended for eitherreception-only or transmissionWreception.

Acknowledgments

I am grateful to all my friends and colleagueswith whom I have interacted professionally. Inparticular, I would like to thank Dr. Xiaoyu Yangand Tsinghua Zheng at Case Western ReserveUniversity (CWRU) for their invaluable discussionson preampliˆer decoupling; Professors Mark A.Griswold (CWRU), Ray F. Lee (New York Univer-sity), and Larry L. Wald (Massachusetts GeneralHospital [MGH]) and Drs. Patrick Ledden (NovaMedical), JeŠ H. Duyn (National Institutes ofHealth), Brian Rutt (Robarts Research Institute),J. Tommy Vaughan (University of Minnesota),Ioannis Panagiotelis (Siemens Medical Solutions),Juergen Kampmeier (Siemens Medical Solutions),and John L. Patrick (Philips Medical Systems)for providing their invaluable picturesWimages;Professor Robert W. Brown (CWRU) for hisinvaluable comments and discussions to improvethe manuscript; and Matthew Finnerty for hisassistance in generating a plot of Fig. 4(b). I amalso grateful for discussions with Drs. JohanOverweg (Philips Research Laboratories, Ham-burg, Germany), Arne Reykowski (InVivo Corpo-ration), and Ralph Oppelt (Siemens CorporateResearch, Germany) at the 2005 ISMRM MREngineering Study Group Workshop on WirelessMRI Coil Technology and with Wayne Dannels(Toshiba Research Institute USA) that guided mydescription on the approach to desirable wirelessdata signal processing. Professors Steve Wright(Texas A & M University), Stuart Crozier (TheUniversity of Queensland), Steve Conolly (Stan-ford University and University of California,Berkeley), Dr. Hubertus Fischer (Siemens MedicalSolutions), and Kazuya Okamoto (Toshiba Medi-cal Systems Corporation) are gratefully acknowl-edged for their useful discussions over the years.Finally, I would like to thank Professor KatsumiKose (Tsukuba University), Dr. Fumiyuki Mitsu-mori (National Institute for Environmental Stud-ies), and Professor Tsuneya Watabe (Editor-in-Chief, Magnetic Resonance in Medical Sciences)for providing me with this wonderful opportunityto write a review article on this RF coil-relatedsubject.

A part of this work was presented at the Four-teenth Scientiˆc Meeting of the ISMRM51 and the34th Annual Scientiˆc Meeting of the JapaneseSociety of Magnetic Resonance in Medicine held atthe Tsukuba International Conference Hall in2006.52

4141RF Coil Designs and Trends

Vol. 6 No. 1, 2007

References

1. Hoult DI. The NMR receiver: a description andanalysis of design. Progress in NMR spectroscopy1978; 12:41–77.

2. Roemer PB, Edelstein WA, Hayes CE, Souza SP,Mueller OM. The NMR phased array. Magn ResonMed 1990; 16:192–225.

3. Hayes CE, Hattes N, Roemer PB. Volume imagingwith MR phased arrays. Magn Reson Med 1991;18:309–319.

4. Fenn AJ, Temme DH, Delaney WP, CourtneyWE. The development of phased-array radar tech-nology. Lincoln Lab J 2000; 12:321–340.

5. Stutzman WL, Thiele GA. Antenna theory anddesign. John Wiley & Sons, 1981; 108.

6. Wright SM, Wald LL. Theory and application ofarray coils in MR spectroscopy. NMR Biomed1997; 10:394–410.

7. Hayes CE, Edelstein WA, Schenck JF, MuellerOM, Eash M. An e‹cient, highly homogeneousradiofrequency coil for whole-body NMR imagingat 1.5T. J Magn Reson 1985; 63:622–628.

8. Kucera J, Lott U. Low noise ampliˆer design, In:Golio M, ed. The RF and microwave handbook.CRC Press, 2001; 548–571.

9. Yang X, Zheng T, Fujita H. TWR switches, baluns,and detuning elements in MRI RF coils, In:ISMRM Fourteenth Scientiˆc Meeting WeekendSyllabus 2006.

10. Sodickson DK, Manning WJ. Simultaneous acqui-sition of spatial harmonics (SMASH): fast imagingwith radiofrequency coil arrays. Magn Reson Med1997; 38:591–603.

11. Pruessmann KP, Weiger M, Scheidegger MB,Boesiger P. SENSE: sensitivity encoding for fastMRI. Magn Reson Med 1999; 42:952–962.

12. Griswold MA, Jakob PM, Heidemann RM, et al.Generalized autocalibrating partially parallel acqui-sitions (GRAPPA). Magn Reson Med 2002; 47:1202–1210.

13. Wiggins GC, Potthast A, Triantafyllou C, et al. A96-channel MRI system with 23- and 90-channelphase array head coil at 1.5T, In: Proceedings ofthe thirteenth annual meeting of the ISMRM.Miami, 2005; 671.

14. Hoult DI, Richards RE. The signal to noise ratioof the nuclear magnetic resonance. J Magn Reson1976; 24:71–85.

15. Froncisz W, Jesmanowicz A, Kneeland JB, HydeJS. Counter rotating current local coils for high-resolution magnetic resonance imaging. MagnReson Med 1986; 3:590–603.

16. Horowitz P, Hill W. In: Chapter 7, The art ofelectronics. Second edition. Cambridge UniversityPress, 1989; 434–435.

17. Li BK. The design and analysis of high frequencyphased array coils for MRI. Ph.D. Thesis. TheUniversity of Queensland, 2006; 84–89.

18. Su S, Zou MX, Murphy-Boesch J. Solenoidal

array coils. Magn Reson Med 2002; 47:794–799.19. Jevtic J. Ladder networks for capacitive decou-

pling in phased-array coils, In: Proceedings ofthe ninth annual meeting of the ISMRM. Glasgow,2001; 17.

20. Wang J, Friedrich A. Antenna arrangement for amagnetic resonance apparatus. U.S. Patent5,708,361.

21. Lee RF, Giaquinto RO, Hardy CJ. Coupling anddecoupling theory and its application to the MRIphased array. Magn Reson Med 2002; 48:203–213.

22. Lee RF, Westgate CR, Weiss RG, Newman DC,Bottomley PA. Planar strip array (PSA) for MRI.Magn Reson Med 2001; 45:673–683.

23. Madore B, Pelc NJ. SMASH and SENSE: experi-mental and numerical comparisons. Magn ResonMed 2001; 45:1103–1111.

24. Herlihy DJ, Larkman DJ, Fujita H, Burl M,Hajnal JV. A 4-channel head coil for SENSEimaging, In: Proceedings of the eighth annualmeeting of the ISMRM. Denver, 2000; 1394.

25. Hajnal JV, Larkman DJ, Herlihy DJ. An arraythat exploits phase for SENSE imaging, In:Proceedings of the eighth annual meeting of theISMRM. Denver, 2000; 1719.

26. Fujita H, Spence DK. A novel 8-channel ``saddle-train'' array coil for abdominal SENSE imaging at1.5T, In: Proceedings of the 10th annual meetingof the ISMRM. Hawaii, 2002; 326.

27. Boskamp E, Blawat L, King KF, Lorbiecki J,Tiwari AB. A 16-channel neurovascular SENSEarray, In: Proceedings of the tenth annual meetingof the ISMRM. Hawaii, 2002; 852.

28. Blawat LR, Hoppel BE, Boskamp EB, Iwadate Y.3-T breast array optimized for parallel imaging, In:Proceedings of the twelfth annual meeting of theISMRM. Kyoto, 2004; 1596.

29. Okamoto K, Takagi M, Ishii M, et al. A novel``smart'' neurovascular array coil system forparallel imaging, In: Proceedings of the twelfthannual meeting of the ISMRM. Kyoto, 2004; 1595.

30. Ohliger MA, Greenman RL, Giaquinto R,McKenzie CA, Wiggins G, Sodickson DK. Con-centric coil arrays for parallel MRI. Magn ResonMed 2005; 54:1248–1260.

31. Spencer D, Akao J, Duensing R, et al. Design of a32-channel cardiac array for parallel imaging, In:Proceedings of the thirteenth annual meeting ofthe ISMRM. Miami, 2005; 911.

32. Lanz T, Kellman, Nittka M, Greiser A, GriswoldM. A 32-channel cardiac array optimized forparallel imaging, In: Proceedings of the fourteenthannual meeting of the ISMRM. Seattle, 2006;2578.

33. Reykowski A, Hemmerlein M, Fischer H. 16-chan-nel headWneck matrix coils for 3 Tesla, In: Pro-ceedings of the thirteenth annual meeting ofthe ISMRM. Miami, 2005; 908.

34. Weiger M, Pruessmann KP, Leussler C,Roschmann P, Boesiger P. Speciˆc coil design for

4242 H. Fujita

Magnetic Resonance in Medical Sciences

SENSE: a six-element cardiac array. Magn ResonMed 2001; 45:495–504.

35. Ohliger MA, Grant AK, Sodickson DK. Ultimateintrinsic signal-to-noise ratio for parallel MRI:electromagnetic ˆeld considerations. Magn ResonMed 2003; 50: 1018–1030.

36. Wiesinger F, De Zanche N, Pruessmann KP.Approaching ultimate SNR with ˆnite coil arrays,In: Proceedings of the thirteenth annual meeting ofthe ISMRM. Miami, 2005; 672.

37. Ledden PJ, Mareyam A, Wang S, van Gelderen P,Duyn JH. 24-channel receive-only array for brainimaging at 7T, In: Proceedings of the fourteenthannual meeting of the ISMRM. Seattle, 2006; 422.

38. Lee RF, Brown R, Mizsei G, et al. Implementationof mode-scanning excitation method with a 16-channel transmitWreceive volume strip array at 7T,In: Proceedings of the fourteenth annual meetingof the ISMRM. Seattle, 2006; 125.

39. Haacke EM, Brown RW, Thompson MR,Venkatesan R. Magnetic resonance imaging: physi-cal principles and sequence design. New York:Wiley-Liss, 1999; 858–859.

40. Yang X, Fujita H, Zheng T, Richmond K, ColwellC. A 3-T QD transmitW8-channel receive hybridknee coil, In: Proceedings of the thirteenth annualmeeting of the ISMRM. Miami, 2005; 907.

41. Vaughan J, DelaBarre L, Snyder C, et al. 9.4-T hu-man imaging: preliminary results, In: Proceedingsof the fourteenth annual meeting of the ISMRM.Seattle, 2006; 529.

42. Katscher U, Boernert P, Leussler C, van den BrinkJS. Transmit SENSE. Magn Reson Med 2003;49:144–150.

43. Zhu Y. Parallel excitation with an array of trans-mit coils. Magn Reson Med 2004; 51:775–784.

44. Kurpad KN, Boskapm EB, Balanis SM. A paralleltransmit volume coil with independent control ofcurrents on the array elements, In: Proceedings ofthe thirteenth annual meeting of the ISMRM.Miami, 2005; 16.

45. Vernickle P, Roschmann P, Findeklee C, et al. An8-channel transmitWreceive body coil for 3T, In:Proceedings of the foureenth annual meeting ofthe ISMRM. Seattle, 2006; 123.

46. Arnulf Oppelt, ed. Imaging systems for medicaldiagnostics. Erlangen: Publicis Corporate Publish-ing, 2005; 830–835.

47. Fischer H, Ladebeck R, Schmand MJ. Challengesand technical solutions of a combined MR PETscanner. Jpn J Magn Reson Med 2006; 26 (Suppl):111.

48. Kuroda K. MR compatible endoscope. NihonKeizai Shinbun 2005; 18 November.

49. Koste GP, Nielsen MC, Tolliver TR, Frey RL,Watkins RD. Optical MR receive coil array inter-connect, In: Proceedings of the thirteenth annualmeeting of the ISMRM. Miami, 2005; 411.

50. Scott G, Yu K. Wireless transponders for RF coils:systems issues, In: Proceedings of the thirteenthannual meeting of the ISMRM. Miami, 2005; 330.

51. Fujita H, Yang X, Zheng T. Array coil types anddesign principles, In: ISMRM Fourteenth ScientiˆcMeeting Weekend Syllabus 2006.

52. Fujita H. New horizons in MR technology: from aperspective of RF coil designs and trends. Jpn JMagn Reson Med 2006; 26 (Suppl): 111.