SCOTT A. MIROWITZ, MD, CONSULTING EDITOR
`
`MAGNETIC
`RESONANCE
`IMAGING
`CLINICS OF NORTH AMERICA
`
`Physics of MR Imaging
`
`J. PAUL FINN, MD, GUEST EDITOR
`
`VOLUME 7
`
`•
`
`NUMBER4
`
`•
`
`NOVEMBER 1999
`
`W.B. SAUNDERS COMPANY
`A Division of Harcourt Brace & Company
`PHILADELPHIA
`LONDON
`TORONTO
`
`MONTREAL
`
`SYDNEY
`
`TOKYO
`
`General Electric Co. 1005 - Page 1
`
`

`
`W.B. SAUNDERS COM PANY
`A Division of HarcoHrt Brace & Company
`
`The Curtis Center • Independence Square West • Philadelphia, Pennsylvania 19106
`http://www .wbsaunders.com
`
`MRI CLINICS OF NORTH AMERICA
`November 1999
`Bditor: Barton Dudlick
`
`Volume 7, Number 4
`ISSN 1064-9689
`Production Editor: Mkhael Pittard
`
`Copyright © 1999 by W.8. Saunders Company. All rights reserve d. No part of this publication may
`in any form or by any means, electronic or mechanical, including
`be reproduced or transmjtted
`photocopy, recording, or any information retrieval system, without permission from the Publisher .
`
`The appearance of the code at the top of the first page of an article in this periodical indicates the
`consent of the W.B. Saunders Company that copies of the article may be made for personal or
`interna l use, or for the personal or internal use of specific clients, for those registered with the Copyright
`Clearance Center, Inc. (222 Rosewood Drive, Danvers, MA 01923; (978) 750-8400; wunv.copyright.com).
`This consent is given on the condition that the copier pay the stated per-copy fee for tltat article
`through the Copyright Clearance Center, Inc. for copying beyond that permitted by Sections 107 or
`108 of the US Copydght Law. This consent does not exten~ to other kinds of copying, such as
`copying for genera l distribution, for advertising or promotional purposes, for creating new collective
`works, or for resale. Absence of the code indicates that the material may not be processed through
`the Copyright Clearance Center, Inc.
`
`The ideas and opinions exp ressed in Magnetic Resonance Imaging Clinics of North America. do not
`necessarily reflect those of the Publisher. The Publisher does not assume any responsibility for any
`injury and I or damage to persons or property arising out of or related to any use of the material
`contained in this periodical. The reader is advised to check the appropriate medical literature and
`the product information currently provided by the manufacturer of each drug to be administered to
`verify the dosage, the method and duration of administration, or contraindications. It is the responsi(cid:173)
`bi_Uty of the treating physician or other health care professional, relyi ng on independent experience
`and knowledge of the patient, to determine drug dosages and the best treatment for the patient.
`Mention of any product in this issue should not be construed as endorsement by the cont ribut ors,
`editors, or the Publisher of the product or manufacturers' claims.
`
`Magnetic Resonance Imaging Clinics of North America (ISSN 1064-9689) is published quarterly by the
`W.B. Saunders Company. Corporate and editorial offices: The Curt is Center, Independence Square
`West, Philadelphia, PA 19106-3399. Accounting and circulation offices: 6277 Sea Harbor Drive,
`Orlando, FL 32887-4800. Periodicals postage paid at Orlando, FL 32862, and addit ional mailing
`offices. Subscription prices are $136.00 per year (US individuals), $173.00 per year (US institutions),
`$153.00 per year (Canad ian individuals) , $209.00 per year (Canadian institutions), $183.00 per year
`(foreign individuals), and $209.00 per year (foreign institutions). Foreign air speed delivery
`is
`included in all Clinics subscription prices. All prices are subject to change without notice. POSTMAS(cid:173)
`TER: Send address changes to Magnetic Resonance Imaging Clinics of Nort/1 America, W.B. Saunders
`Company, Per iodicals Fulfillment, Orlando, FL 32887-4800. Cus tomer Service: 1-800-654-2452 (US).
`From outside of the US, call 1-407-345-4000.
`
`Magnetic Resonance Imaging Clinics of North America is covered in the RSNA Index of Imaging Literature,
`Index Medic11S, MEDLINE, and EMBASE!Excerta Medica.
`
`Printed in the United States of America.
`
`General Electric Co. 1005 - Page 2
`
`

`
`PHYSICS OF MR IMAGING
`
`1064-9689/99 $8.00 + .00
`
`OVERVIEW OF MR IMAGING PULSE
`SEQUENCES
`
`John P. Mugler ill, PhD
`
`The tremendous contrast flexibility of MR
`imaging results from the literally unlimited
`variety of pulse-sequence
`techniques. Even
`though MR imag ing is, in many ways, a ma(cid:173)
`ture imaging modality, pulse sequences con(cid:173)
`tinue to be developed that provide significant
`improvements for existing applications or im(cid:173)
`portant new capabilities. Nonetheless, every
`pulse sequence must accomplish two basic
`tasks. First, transverse magnetization must be
`created using one or more radiofrequency
`(RF) pulses, and this magnetization must then
`be encoded using gradient pulses so that the
`spatial positions of the tissues can be deter(cid:173)
`mined to form the image. Second, the desired
`contrast between
`tissues must be created
`based on the time of application and duration
`of the various RF and gradient pulses. In
`this article, after a review of general pu lse(cid:173)
`sequence properties, the features of the tech(cid:173)
`niques most commonly used for clinical MR
`imaging are discussed. In particu lar, the man(cid:173)
`ner in which each technique generates and
`encodes transverse magnetization
`and the
`typical forms of image contrast available from
`each technique are emphasized.
`
`GENERALPULS&SEQUENCE
`PROPERTIES
`
`The Pulse-Sequence Timing Diagram
`
`the im(cid:173)
`A useful method for conveying
`portant characteristics of a pulse sequence is
`a graphical representation called the pulse(cid:173)
`seq uence timing diagram. These diagrams
`will be used extensively when specific tech(cid:173)
`niques are described. The major features of
`timing diagrams will be explained using a
`single spin-echo pulse sequence as an exam(cid:173)
`ple (Fig. 1).
`In general, the timing diagram depicts the
`temporal sequence of the RF, gradient, and
`data-sampling events in the pulse sequence.
`The first line, labeled RF, illustrates the time
`of application and the waveforms for the RF
`pulses, in this case 90° (excitation) and 180°
`(refocusing) pulses. The next three lines de(cid:173)
`pict the time of application and the wave(cid:173)
`forms for gradients applied along three mutu(cid:173)
`ally orthogonal axes. These are labeled Gss,
`GrEJ and GRo, for the section-select, phase(cid:173)
`encoding, and readout gradients, respectively,
`although other labels, for example Gx, G~ Gz,
`
`From the Departm ents of Radiology and Biomedi cal Engin eering, University of Virginia School of Medi cine, Charlottes(cid:173)
`ville, Virginia
`
`MRI CLINICS OF NORTH AMERICA
`
`VOLUME 7 • NUMBER 4 • NOVEMBER 1999
`
`661
`
`General Electric Co. 1005 - Page 3
`
`

`
`662
`
`MUGLER
`
`RF
`
`Gss
`
`Figure 1. Pulse-sequence timing for a single spin-echo technique. RF =
`radiofrequency pulses; Gss = section-select gradient; GPe = phase(cid:173)
`encoding gradient; GRo = readout gradient; DAO = data acquisition; N20
`, = number of times that the basic pulse-sequence timing is repeated,
`which, in this case, equals the number of in-plane phase-encoding lines;
`TR = repetition time; and TE = echo time.
`
`\
`
`can be used to indicate the assignment of the
`gradients for a particular section orientation.
`For the two-dimensional (20) acquisition de(cid:173)
`picted in Figure 1, section-select gradients are
`applied in synchrony with the RF pulses, thus
`spatially localizing their effects, or in other
`words selecting the section of interest. The
`other gradient waveforms, applied along the
`phase-encoding and readout axes, spatially
`encode the magnetization within this section
`along the remaining two directions. The gra(cid:173)
`dient waveform on the phase-encoding axis,
`containing a series of closely spaced hori(cid:173)
`zontal lines, is the graphical representation of
`a gradient table: the strength of the gradient
`is stepped through a series of values as the
`basic pulse-sequence
`timing is repeated
`to
`collect the data required to form the image.
`With each repetition, another line of data is
`collected. The fifth line in the timing diagram
`shows when the MR signals occur. Although
`some pulse sequences generate a number of
`distinct signals, typically only the signals of
`interest are depicted. The last line of the tim(cid:173)
`ing diagram, labeled OAQ for data acquisi(cid:173)
`tion, illustrates the time period during which
`the data are collected. This line is sometimes
`labeled ADC, for analog-to-digital convertor.
`
`It is not uncommon for either or both of the
`signal and DAQ lines to be omitted. Finally,
`the square brackets around all of the timing
`events with the N 20 label in the lower right
`comer indicate that these events are repeated
`N20 times.
`
`Modes of Spatial Encoding
`
`MR images are typically acquired either as
`a set of 20 sections or as a three-dimensional
`(30) volume (Fig. 2). In a 30 acquisition, an
`additiona l phase-encoding gradient table is
`applied along the third dimension to yield a
`set of contiguous sections within the volume .
`These sections are often referred to as parti(cid:173)
`tions to distinguish them from image sections
`derived from a fundamentally 20 acquisition.
`The advantages of a 30 acquisition include
`the availability of thin, contiguous sections
`that can be used to obtain high -resolution
`images of arbitrary orientations through post(cid:173)
`processing11, 72. 112, and, for pulse-sequence pa(cid:173)
`rameter values that are equivalent in the 20
`and 30 acquisitions, an increase in the signal(cid:173)
`to-noise ratio (SNR) by a factor equal to the
`square root of the number of partitions. 106 A
`
`General Electric Co. 1005 - Page 4
`
`

`
`OVERVIEW OF MR IMAGING PULSE SEQUENCES
`
`663
`
`20 sections
`
`30 single-slab
`
`30 multi-slab
`
`Figure 2. The three basic modes of spatial encoding used in MR imaging: two-dimensional
`(20) sections, three-dimensional (30) single-slab, and 30 multislab. This example illustrates
`the possibilities for acquiring axial images of the brain. Acquisitions using 20 sections or 30
`multislab often employ a gap between the sections or slabs, as shown, to increase the
`anatomical coverage and to decrease crosstalk (significant degree of interference) effects,
`although the use of contiguous sections or slabs (i.e., no gaps) is appropriate for some
`applications.
`
`hybrid of the 20 and 30 acquisition modes,
`called multislab
`30,
`is also sometimes
`133 For this acquisition a set of slabs is
`used. 103
`•
`acquired, and each slab in tum is phase en(cid:173)
`coded along the third dimension to yield a
`set of contiguous partitions (Fig. 2). With ref(cid:173)
`erence to this mode, single -volume 30 acqui(cid:173)
`sition is sometimes referred to as single-slab
`30.
`The 20 mode of acquisition can be subdi(cid:173)
`vided into two important forms: sequential
`section and multisection. 19 In a sequential(cid:173)
`section acquisition, which is analogous to the
`operation of conventional computed
`tomog(cid:173)
`raphy (CT), all of the data required for a
`given image are collected before proceeding
`to the next image. A multisection acquisition,
`however, collects only a portion of the re(cid:173)
`quired data for a given image, typically one
`or several phase-encoding
`lines, before col(cid:173)
`lecting the corresponding data for each of the
`other images. This process is repeated until
`all of the data for all of the images have been
`
`collected (Fig. 3). Obviously, a given image
`can be acquired more rapidly using the se(cid:173)
`quential-section method, and this method is
`typically used when short acquisition times
`are critical, for example, in some implementa (cid:173)
`tions of MR angiography or for imaging rap (cid:173)
`idly moving structures. On the other hand, a
`multisection
`acquisition
`is typically used
`when a relatively long time period (on the
`order of 1 second) is required between the
`interrogations of a particular section to permit
`Tl recovery to occur and thereby generate the
`desired image contrast.
`Another practical aspect of sequential-section
`and multisection acquisitions is the temporal
`order in which the data for the sections are
`collected. This may be important, for exam(cid:173)
`ple, if there is blood flow perpendicular to the
`sections, 2• 7 or if there is a significant degree of
`(crosstalk) between sections. 71• 74
`interference
`The most common acquisition orders are con(cid:173)
`secutive, beginning at either end of the image
`set, and interleaved, for example collecting all
`
`General Electric Co. 1005 - Page 5
`
`

`
`SECTION 1
`
`SECTION 2
`
`SECTION 3
`
`664
`
`MU CLER
`
`DATA
`SEGMENT'
`
`A
`
`2
`
`3
`
`SEQUENCE REPE:TIONEJ
`
`SEQUENCE
`
`REPE:TIO~Ej ~ ~ EJ § g Ej ~ ~
`
`Figure 3. The order of data acqu isition for 20 multisect ion and sequential-sect ion imaging. A
`hypothetica l case is illustrated ' where three images are to be acqu ired, and the k-space data
`corresponding to each image is com posed of three segments . One data segment of an image section
`is acquired during each repetition of the pulse sequence , thus nine repetitions are required to collect
`the data for all three images. A, The k-space data layout corresponding to the three images. B, The
`data ac,quisition order for 20 multisection imaging. C, The data acquisi tion orde r for 20 sequential(cid:173)
`section imaging.
`
`odd-numbered sections followed by all even(cid:173)
`numbered sections, although arbitrary acqui(cid:173)
`sition orders can be used. As a final point,
`the sequential-section and multisection con(cid:173)
`cepts are also applicable to multislab 30 ac(cid:173)
`quisitions.
`Acquisitiorn modes other than the 20 and
`30 forms described above are sometimes
`used. For example, a thin column of tissue
`can be imaged by using either two RF pulses
`that excite intersecting planes 30 or a single 20
`RF pulse .109 Applications of this mode include
`analysis of cardiac function 110 and rapid dilfu(cid:173)
`sion imaging. 8 1
`
`Steady-Statei Versus Transient
`Acquisitions :
`
`Many of the MR pulse sequences in clinical
`use operate in the steady-state mode in which
`severa l repeti .tions of the basic pulse-sequence
`
`timing are required to collect the necessary
`data (see Fig. 1). These repetitions are spaced
`uniformly in time, so that the amplitude and
`phase of the magnetization when the data are
`collected are the same in all repetitions, aside
`from the effects of the phase-encoding gradi(cid:173)
`ent. Anything that disrupts this steady state
`may result in undesired variations in the am(cid:173)
`plitude or phase of the magnetization and, as
`a consequence, may produce image artifacts.
`Such disruptions
`include heart rate-depen(cid:173)
`in the repetition
`time (TR)
`dent variations
`during electrocardiogram (ECG)-triggered ac(cid:173)
`quisitions.
`In contrast with a steady -state acquisition,
`some pulse sequences operate in a transient
`mode in which the amplitude or phase of the
`magnetization varies throughout
`the cou1·se
`of data collection. Such methods of acquisi(cid:173)
`tion are commonly referred to as single-shot
`methods.
`Several important pulse sequences operate
`
`General Electric Co. 1005 - Page 6
`
`

`
`OVERVIEW OF MR lMAGlNG PULSE SEQUENCES
`
`665
`
`in a mode that is a hybrid between the pure
`steady -state and transient acquisitions.
`In
`these techniques,
`the basic pulse-sequence
`timing is repeated several times, just as in a
`steady-state acquisition. During each repeti(cid:173)
`tion, however, the data collection for a given
`image spans a period during which the am(cid:173)
`plitude or phase of the magnetization varies
`substantially, for example because of relax(cid:173)
`ation or the application of RF pulse:s. Hence,
`the data that are collected reflect these varia(cid:173)
`tions. These acquisition methods are often re(cid:173)
`ferred
`to as segmented methods. 25 When
`pulse sequences are described in this article,
`the temporal mode of the acquisition is desig(cid:173)
`single-shot .. or seg(cid:173)
`nated as steady-state,
`mented.
`The total time required to genera 1te an im(cid:173)
`age or set of images with a selected contrast
`behavior is often one of the most important
`characteristics of the pulse sequence. For most
`steady-state and segmented
`techniques,
`the
`imaging time can be calculated from
`Imaging Time = TR-NREl'·NAc
`where NREP is the number of times that the
`basic pulse-sequence
`timing is repeated
`to
`permit collection of the data required to form
`the image, and N AC is the number of times
`that the complete data collection process is
`repeated to perform averaging for reduction
`of the noise level. The term NAo (number of
`acquisitions), is also commonly denoted NEX,
`for number of excitations. For many steady(cid:173)
`state pulse sequences, NREP is equal to N20 N30 ,
`where N20 is the number of in-plallle phase(cid:173)
`encoding lines, and NJo is the number of par(cid:173)
`titions within the volume for single-slab 30
`or within each subvolume for multislab 30.
`For 20 acquisitions, N30
`is set equal
`to 1.
`Segmented acquisitions typically acquire sev(cid:173)
`eral phase-encoding
`lines for each repetition
`of the basic pulse-sequence timing, and hence
`the value of NREP must be reduced by an
`appropriate factor. For example, if the num(cid:173)
`ber of phase -encoding lines acquired per rep(cid:173)
`etition is denoted ETL (for echo train length),
`then NREP becomes (N2oNJo)/ ETL.
`A further distinction concerning the tempo(cid:173)
`ral characteristics of a pulse sequence can be
`made by whether the tasks of spatial encod-
`
`(1)
`
`ing and contrast generation are intimately
`coupled or separable. In many steady-state
`pulse sequences, the same set of RF and gra(cid:173)
`dient pulses that generate the contrast also
`spatially encode the magnetization. On the
`other hand, the contrast generation in many
`transient pulse sequences is performed by a
`series of RF and gradient pulses that is dis(cid:173)
`tinct from those that perform spatial encod (cid:173)
`ing. Pulse sequences that use a separate con(cid:173)
`trast preparation
`are often referred
`to as
`49• 93 Of
`magnetization -prepared
`techniques.
`course, the contrast can also be generated in
`a hybrid fashion, using the combined effects
`of a magnetization preparation and the por(cid:173)
`tion of the pulse sequence that performs spa(cid:173)
`tial encoding.
`
`The Effects of Radiofrequency
`Pulses on Magnetization
`
`To facilitate discussion of the various pulse
`sequences, it is useful to review the ways in
`which an RF pulse can affect the magnetiza(cid:173)
`tion. 130 Consider an RF pulse with a flip angle
`e applied to a longitudinal component of the
`magnetization (M2 ). Just after this RF pulse,
`a fraction of the original magnetization
`(Mzcos0) remains along the longitudinal axis,
`and transverse magnetization
`(of amplitude
`M2sin0) is also created. Note that transverse
`magnetization
`is created for any value of 0
`except for an integer multiple of 180°.
`The effects of an RF pulse on a transverse
`component of the magnetization
`(MXY) are
`more complex. Just after the RF pulse, a frac(cid:173)
`tion of the original magnetization [Mxycos2(0/
`2)] remains
`in the
`transverse
`plane, un(cid:173)
`affected by the application
`of the pulse.
`Another
`portion
`of
`the magnetization
`(Mxvsin 2(0 /2)] remains
`in the
`transverse
`plane, but the phase state of this magnetiza(cid:173)
`tion is reversed (conjugated) by the applica(cid:173)
`tion of the pulse. It is this latter portion of
`the magnetization
`that can evolve to form a
`spin echo. Note that phase-reversed
`trans(cid:173)
`verse magnetization
`is created for any value
`of O except for an integer multiple of 360°. In
`addition
`to producing
`two transverse com-
`
`General Electric Co. 1005 - Page 7
`
`

`
`666
`
`MUGLER
`
`ponents, an RF pulse applied to transverse
`magnetization
`also results in two distinct
`components of longitudinal magnetization.
`Both components have the same amplitude
`[Mxv(sin0)/2], and they both retain the phase
`inform ation of the transverse magnetization.
`The difference between
`th ese two compo(cid:173)
`nents is that the phase information
`in one
`component
`is reversed, analogous
`to the
`phase-reversed transverse magnetization, and
`the phase information in the other component
`is not. If the phase-reversed
`longitudinal
`magnetization
`is converted
`into transverse
`magnetization by a subsequent RF pulse, it
`can evolve to form a stimu lated echo.37
`Radiofrequen cy pulses are often given a
`qualitative label based on their desired effect.
`For example, pulses that have the purpose of
`creating transverse magnetization are gener(cid:173)
`ally called excitation pulses, and pul ses that
`have the purpose of reversing the phase state
`of transverse magnetization
`are generally
`called refocusing pulses. Inversion pulses are
`intended
`to convert positive
`longitudinal
`magnetization
`into negative
`longitudinal
`magnetization without creating any trans(cid:173)
`verse component. The flip angle of such a
`pulse is thus 180°. Note, how ever, as can be
`deduced
`from the relationships discussed
`previously, that any excitation pulse poten(cid:173)
`tially performs some degree of refocusing,
`and any refocusing pulse (except integer mul(cid:173)
`tiples of 180°) potentially performs some de(cid:173)
`gree of excitation. This behavior leads to a
`large number of higher-order echoes when
`more than a small number of RF pulses are
`applied in a time period that is short com(cid:173)
`pared with the tissue relaxation times.58 In
`this regard, pulse seq uences that employ
`many closely spaced RF pul ses must be care(cid:173)
`fully designed to ensure that only the desired
`echo signals contribute to the image.
`
`k-Space Trajectories and Filters
`
`The MR signals recorded from the patient
`are actually the spatia l-frequency space, or k(cid:173)
`space, description of the image. Therefore,
`the desired image is formed by a Fourier
`transform of these k-space data. An important
`
`characteristic of many pulse sequences is the
`temporal order, or arrangement, used to col(cid:173)
`lect the k-space data. 9• 80•
`124 This order of data
`collection is commonly referred to as the k(cid:173)
`space trajectory. Foir example, in conventional
`spin-echo imaging, the k-space data are typi(cid:173)
`ca 11 y collected sequentially,
`line by line,
`moving from one ,extreme of k-space to the
`opposite extreme. In contrast, fast imaging
`techniques often use other types of k-space
`trajectories, such as spirals. 91 In principle, any
`form of k-space trajectory can be used if two
`conditions are met. First, the region of k-space
`that is covered must be sufficiently large to
`describe the anatomy of interest with the de(cid:173)
`sired spati al resolultion. Second, the minimum
`spacing between data samples must be suffi(cid:173)
`ciently small
`to prevent aliasing
`(wrap(cid:173)
`around) artifacts. Major factors affecting the
`selection of the k-space traje_ctory include the
`speed of image acquisition, ·the desired con(cid:173)
`trast, and the level of artifacts.
`The relationship of the k-space trajectory
`to the image contrast behavior and potential
`artifacts can be understood by reviewing
`some of the basic properties of k-space. The
`data in the central portion of k-space, that is,
`the low spatiaJ-fre,quency data, represent the
`gross structure and contrast
`in the image.
`Conversely, the data in the outer portions
`of k-space, that is, the high spatial-frequency
`data, represent the detailed structure and con(cid:173)
`trast in the image. The position of the k-space
`data relative to the center corresponds to the
`orientation of the structures that the data rep(cid:173)
`resent in the image. These principles are illus(cid:173)
`trated in Figure 4. For example, in Figure 4C
`only k-space data for high-spatial frequencies
`are present. Thus, only features with rapidly
`changing intensity values (that is, edges, par(cid:173)
`ticularly in the vicinity of the bright subcuta(cid:173)
`neous fat) are expected to be seen in the cor(cid:173)
`responding image .. Further, only k-space data
`in the vicinity of the horizontal axis are in(cid:173)
`cluded. Therefore, only edges oriented paral(cid:173)
`lel to the vertical direction are visible.
`Recall, as discussed previously, that the am(cid:173)
`plitude and phase: of the magnetization gen(cid:173)
`erally vary thrornghout the course of data
`collection for a single-shot or segmented ac(cid:173)
`quisition and may also vary in a steady-
`
`General Electric Co. 1005 - Page 8
`
`

`
`OVERVIEW OF MR IMAGING PULSE SEQUENCES
`
`667
`
`~lgure 4. The relationship between the data location in k-space and the resulting features in the
`image. The k-space data (upper row) and corresponding images (lower row) are shown for (A) the
`complete 256 x 256 data set, (B) the central 64 x 64 data points, (C) two 96 x 32 data blocks
`centered on the horizontal axis, anc! (D) two 32 x 96 data blocks centered on the vertical axis. The
`k-space data is displayed using a k>garithmic intensity scale to provide an improved visualization of
`the high spatial-frequency compone,nts.
`
`state acquisition because of effects such as
`motion. The extent of these variations de(cid:173)
`pends on the details of the pulse sequence
`and the physical properties, such as the relax(cid:173)
`ation times, of the anatomy. Co:nsidering
`these potential signal variations and the rela(cid:173)
`tionships between k-space and image, it is
`obvious that the relative amplitude .and phase
`of the data may vary as a function of the
`location in k-space. This variation as a func(cid:173)
`tion of spatial frequency is often ca:Ued the k(cid:173)
`space filter of the acquisition. Depcmding on
`the form of this filter, as determined by the
`k-space trajectory, image artifacts such as
`blurring, edge enhancement, or gho,sting may
`result. Figure 5 shows an example ,of k-space
`filter effects that could be caused by T2 relax(cid:173)
`ation in a segmented acquisition. In Figure
`SB, the k-space filter attenuates high spatial(cid:173)
`frequency components along one direction.
`Because these data correspond
`to detailed
`structure, blurring is seen in the correspond(cid:173)
`ing image along the same directio,n. On the
`other hand, the k-space filter in Figure SC
`enhances the high spatia l-frequency compo(cid:173)
`nents, and the corresponding image demon(cid:173)
`strates edge enhancement.
`
`Criteria for Pulse Sequence
`Evaluation
`
`To choose the best pulse sequence for a
`particular imaging application, it is necessary
`to evaluate the pertinent features of the avail(cid:173)
`able techniques. The major features of impor(cid:173)
`tance include (1) the speed of image acquisi(cid:173)
`tion; (2) the SNR, which is affected by the
`signa l levels and the acquisition (receiver)
`bandwidth; (3) the available image contrasts,
`for example Tl, T2, T2*, proton-density, or
`diffusion weighting; (4) the sensitivity to off(cid:173)
`resonance signals caused by static-field inho(cid:173)
`mogeneities, magnetic-susceptibility
`inter(cid:173)
`faces, or chemical shift; (5) the sensitivity to
`flow or motion; (6) the sensitivity to hardware
`performance; and (7) the complexity of the
`reconstruction algorithms . The descriptions of
`the various pulse sequences highlight the fea(cid:173)
`tures that are a particular advantage or disad(cid:173)
`vantage of each technique.
`
`Pulse Sequence Acronyms
`
`It is very common for MR techniques to
`be referred to by acronyms that are typically
`
`General Electric Co. 1005 - Page 9
`
`

`
`668
`
`MUGLER
`
`A
`
`B
`
`c
`
`SPATl !AL FREQUENCY
`
`SPATIAL FREQU ENCY
`
`SPATIAL FREQU ENCY
`
`Figure 5. Simulation of k-space filter effects. The filter functions (upper row), filtered k·space data
`(middle row), and the corresponding images (bottom row) are shown for (A) an ideal, uniform k(cid:173)
`space filte r, and (B) an attenuation or (C) an enhancement of the high spatial-frequency components
`along one axis of k-space. The images in (B) and (C) were generated by applying the respective,
`simulated k-space filters to the experimentally measured k-space data for (A). The k-spaoe data is
`displayed! using a logarithmic intensity scale to provide an improved visualization of the high spatial(cid:173)
`frequency components.
`
`derived from major structural or operational
`features of the pulse sequence . This discus(cid:173)
`sion of the techniques most commonly used
`for clinical MR imaging uses generic terms,
`when possible, to describe the features and
`functions of each technique but also provides
`some of the more common acronyms. 28 Note,
`hovvever, that although pulse-sequence acro(cid:173)
`nyms can serve as useful abbreviations, they
`can also be obstacles for the novice MR user
`who is trying to learn the array of available
`techniques and how they work. This is partic(cid:173)
`ularly true for many of the acronyms used by
`the equipment manufacturers, because totally
`
`different names are often used for basically
`identical techniques. Most clinically applica(cid:173)
`ble pulse -sequence
`techniques are available
`on the MR platforms of all manufacturers,
`and, even though the names are different, the
`underlying physical principles of any given
`technique are the same, regardless of the im(cid:173)
`aging platform.
`
`SPIN-ECHO-BASED PULSE
`SEQUENCES
`
`The review of spin -echo (SE) imaging be(cid:173)
`gins with the basic single- and double-SE
`
`General Electric Co. 1005 - Page 10
`
`

`
`OVERVTEW OF MR IMAGING PULSE SEQUENCES
`
`669
`
`puJse sequences that are the foundation for
`many of the more complex SE-based meth(cid:173)
`ods. Next, a family of rapid SE techniques are
`presented and followed by a discussion of the
`inversion-recovery contrast preparation
`that
`is used with a variety of SE-based methods.
`Finally, a class of ultra-fast SE techniques is
`described.
`
`Convention al Spin-Echo Imaging
`
`Only a few years ago, most clinical MR
`imaging was performed using what is now
`called conventional SE imaging . The word
`conven tional was added to the name recently
`to distinguish this form of SE-based imaging,
`in which one phase-encoding line of k-space
`data is collected per image with each excita(cid:173)
`tion RF pulse, from the more rapid methods
`that collect several phase-encoding lines with
`each excitation pulse. Even with the advent
`of faster SE-based techniques, however, con(cid:173)
`ventional SE is still widely used beca~se of
`its well understood contrast characteristics
`and its resistance to image artifacts from a
`variety of sources such as RF or static-field
`inhomogeneities.
`The timing diagram for one of the work(cid:173)
`horses of conventional SE imaging, the dou-
`
`ble-echo puJse sequence, is shown in Figure
`6. As illustrated in the figure, the RF pulse
`scheme includes an excitation RF pulse, typi(cid:173)
`cally using a flip angle of 90°, followed by
`one or more refocusing RF pulses, typically
`using a flip angle of 180°. The important fea(cid:173)
`ture of the spatial -encoding scheme is that all
`echoes collected in any given repetition are
`encoded the same. Thus, for multiple refocus(cid:173)
`ing RF pulses, the echo signals recorded after
`each pulse are used to generate images with
`different T2 weightings. That is, each echo
`time (TE) corresponds
`to a separate image.
`Conventional SE imaging is a steady -state
`technique and is used almost exclusively to
`perform 20 multisection acquisitions. Its typi(cid:173)
`cal k-space trajectory is simp le, collecting data
`sequentially, line by line, moving from one
`extreme of k-space to the opposite extreme.
`Aside from the effects of motion, no k-space
`filtering is inherent to a conventional SE ac(cid:173)
`quisition.
`Although the most common implementa(cid:173)
`tions employ a flip angle of 90° for the excita(cid:173)
`tion RF pulse, flip angles other than 90° are
`sometimes used. 6, 21, 89 For example, in a sin(cid:173)
`gle-echo Tl-weighted pulse sequence the ex(cid:173)
`citation flip angle can be decreased to yield
`increased Tl weighting at the expense of de(cid:173)
`creased SNR. In a double-echo T2-weighted
`
`1so·
`go·
`
`1so·
`
`Figu re 6. Pulse-sequence timing for a double spin-echo technique. RF
`= radiofrequency pulses; Gss = section-select gradient ; GPe = phase(cid:173)
`encoding gradient; GRo = readout gradient; DAO = data acquisition; and
`N20 = number of in-plane phase-encoding lines.
`
`General Electric Co. 1005 - Page 11
`
`

`
`670
`
`MUGLER
`
`pulse sequence the excitation flip angle can
`be decreased to gener