INTRAOPERATIVE MRI DEVELOPMENTSCONTENTS

Preface xiChristopher Nimsky and Rudolf Fahlbusch

Basic Principles of Magnetic Resonance Imaging 1Wendell A. Gibby

We have come full circle from spinning quarks to three-dimensional (3D) medicalimages. The bulk of MRI is now performed using slice-selective gradients, where radio-frequency energy is applied to excite the hydrogen nuclei. By stepping a phase-encodinggradient during each repetition time and using a frequency-encoding gradient as thedata are sampled, the 3D human object can be reduced to many individual points or vox-els. By acquiring multiple slices at once, the time efficiency of imaging can be vastlyimproved. Many newer strategies use variations of this technique to acquire multiplelines of data during a single echo, enshrining spin warp imaging as the most importantmethod of signal acquisition for MRI.

QuasieReal-Time Neurosurgery Support by MRI Processing via GridComputing 65Heiko Lippmann and Frithjof Kruggel

In this article, a parallel image processing tool chain to correct preoperative functionalMRI data with respect to the brain shift phenomenon based on intraoperative MRI scansof the patients head is introduced. For this purpose, nonrigid image registration of ana-tomic intraoperative MRI based on a fluid dynamical model is performed to gain a three-dimensional displacement field reflecting deformations of the brain tissue. To achieve aclinically acceptable run time, the use of grid computing is aimed at intensive computingon a remote personal computer cluster. To obtain a secure and reliable computation ser-vice over the Internet, a newly developed European grid technology is used.

Functional MRI Localizing in the Cerebellum 77Wolfgang Grodd, Ernst Hulsmann, and Hermann Ackermann

Mapping of cerebellar function by functional MRI now enables us not only to re-estab-lish older anatomic findings of somatotopic representations but to gain new insights inthe function of the cerebellum and its intimate relations of cerebral regions to servingsensorimotor function, sensory discrimination, and cognitive processing. Consequently,

it will change our understanding of neurologic and psychologic failures in patients withinborn errors or neurodegenerative diseases or after neurosurgical procedures.

VOLUME 16 NUMBER 1 JANUARY 2005 v

Proton Magnetic Resonance Spectroscopic Imaging in Brain TumorDiagnosis 101Stephen Gruber, Andreas Stadlbauer, Vladimir Mlynarik, Brigitte Gatterbauer,Karl Roessler, and Ewald MoserThe current state of standard tumor diagnostics using contrast-enhanced MRI andbiopsy is assessed in this review, and the progress of proton magnetic resonance spectro-scopy (MRS) over the last 15 years is discussed. We summarize MRS basics and describea typical magnetic resonance session for noninvasive routine tumor diagnostics at 1.5 T,including two-dimensional magnetic resonance spectroscopic imaging (MRSI). The re-sults that can be obtained from such procedures are illustrated with clinical examples.Attention is turned to cutting-edge methodologic and clinical research at 3 T, with exam-ples using high-resolution or very short echo-time three-dimensional MRSI. The currentstatus and limitations in proton MRSI are discussed, and we look to the potential offaster data collection and even higher field strength.

Diffusion Tensor Magnetic Resonance Imaging of Brain Tumors 115A. Gregory Sorensen

Diffusion tensor imaging (DTI) appears to offer the possibility of adding important in-formation to aid in presurgical planning. Although experience is limited, DTI seems toprovide useful local information about the structures near the tumor. In the future,DTI may provide an improved way to monitor intraoperative surgical procedures aswell as their effects. Evaluation of the response to treatment with chemotherapy and ra-diation therapy may also become possible. Although DTI has some limitations, its activeinvestigation and further study are clearly warranted.

A Low-Field Intraoperative MRI System for Glioma Surgery: Is itWorthwhile? 135Dennis S. Oh and Peter M. Black

Intraoperative MRI has proven to be a crucial tool in glioma surgery. Over the past dec-ade, more than two thirds of the 871 operations done in our facility have been on glio-mas. The surgical challenges related to brain shift, discernment of tumor margins,eloquent cortex, and completeness of resection are met by the unprecedented capabilitiesof intraoperative MRI. It allows precise localization of tumor margins and neural struc-tures and provides updated information on the progress of surgery. The result is thor-ough tumor resection without critical injury to important areas of the brain. Asintraoperative MRI continues to expand its presence, it is likely to become a standardof care for many glioma cases.

Intraoperative Magnetic Resonance Imaging at 0.12 T: Is it Enough? 143Michael Schulder, Jeffrey Catrambone, and Peter W. Carmel

Compact imagers for intraoperative MRI (iMRI) designed for use in a regular neurosur-gical operating room (OR) are an attractive alternative to modifying a diagnostic MRI(dMRI) suite for surgery or altering an OR to accommodate dMRI. The PoleStar N-10iMRI system incorporates a 0.12-T magnet and was fashioned as a tool for intracranialneurosurgery. In our experience, this system proved to be a valuable aid for a wide vari-ety of surgery, mostly for intracranial tumors. Expansion of this compact unit to a unitwith a 0.15-T magnet has recently been accomplished, addressing some of the limitationsof the previous device. We discuss the pros and cons of surgery with these low-fieldcompact iMRI systems.

vi CONTENTS

Adaptation of a Standard Low-Field (0.3-T) System to the Operating RoomFocus: Pituitary Adenomas 155Borimir J. Darakchiev, John M. Tew, Jr, and Ronald E. WarnickIntraoperative MRI (iMRI) is a reliable and safe tool to monitor the extent of resectionand to avoid complications in the transsphenoidal surgical approach for pituitary tu-mors. The best indication for its application in transsphenoidal surgery is for patientswith pituitary macroadenomas with suprasellar extension. The low-field 0.3-T magnethas a diagnostic imaging quality that provides surgeons with good intraoperative detailof the anatomic relations in the sellar region. In our experience, iMRI provided a distinctbenefit in planned subtotal resection for invasive macroadenomas that compress the op-tic chiasm and in planned gross total resection for noninvasive tumors. The iMRI designadopted at our center includes important features, such as the use of ferromagnetic sur-gical instruments, elimination of patient transportation, and capability as a sharedresource, that allow multipurpose diagnostic use and increased cost-effectiveness.

1.5 T: Spectroscopy-Supported Brain Biopsy 165Walter A. Hall and Charles L. Truwit

The technique for performing brain biopsy has evolved significantly over the last threedecades. Intraoperative MRI guidance has enhanced the diagnostic rate for brain biopsyby now allowing neurosurgeons to compensate for brain shift while performing the pro-cedure in nearereal time. The development of a trajectory guide enables the neurosur-geon to determine a safe and accurate path for intraoperative MRI-guided brainbiopsy and to secure the position of the needle within the target tissue. Magnetic reso-nance spectroscopy (MRS) has been used to help distinguish recurrent brain tumor fromthe effect of previous treatments by measuring specific metabolites within the area ofconcern. Combining the use of a trajectory guide with MRS should enhance the diagnos-tic yield for MRI-guided brain biopsy.

Epilepsy Surgery with Intraoperative MRI at 1.5 T 173John J. Kelly, Walter J. Hader, S. Terry Myles, and Garnette R. Sutherland

Monitoring surgical procedures for the treatment of epilepsy is a relatively new applica-tion of intraoperative MRI (iMRI). At the University of Calgary, an iMRI system based ona moveable 1.5-T magnet has been developed and applied to the surgical management ofrefractory epilepsy. Seventy patients have been prospectively evaluated during treat-ment in this patient-focused environment. This article reviews the experience and pro-vides insight and direction for future procedures with the goal of continuing theadvancement of epilepsy surgery.

1.5 T: Intraoperative Imaging Beyond Standard Anatomic Imaging 185Christopher Nimsky, Oliver Ganslandt, and Rudolf Fahlbusch

Intraoperative high-field MRI with integrated microscope-based neuronavigation is asafe and reliable technique providing immediate intraoperative quality control. Major in-dications are pituitary tumor, glioma, and epilepsy surgery. Intraoperative high-fieldMRI provides intraoperative anatomic images at high quality that are up to the standardof pre- and postoperative neuroradiologic imaging. Compared with previous low-fieldMRI systems used for intraoperative imaging, not only is the image quality is clearlysuperior but the imaging spectrum is much wider and the intraoperative work flow isimproved. Furthermore, high-field MRI offers various modalities beyond standard ana-tomic imaging, such as magnetic resonance spectroscopy, diffusion tensor imaging, andfunctional MRI.

CONTENTS vii

Future Perspectives for Intraoperative MRI 201Ferenc A. Jolesz

MRI-guided neurosurgery not only represents a technical challenge but a transformationfrom conventional hand-eye coordination to interactive navigational operations. In thefuture, multimodality-based images will be merged into a single model, in which anat-omy and pathologic changes are at once distinguished and integrated into the sameintuitive framework.

Index 215viii CONTENTS

quality control.

In this issue of Neurosurgery Clinics of NorthAmerica, we focus on current MRI developmentswith an impact on intraoperative use in neurosur-

gery and on the intraoperative application of MRItechnology. This issue compiles the contributionsfrom a variety of experts in their respective

specialties.In the rst part, a general overview of MRI

techniques is followed by focusing on currentdevelopments with a distinct impact on intra-

operative application, ranging from functionalimaging with fMRI, to investigation of metabo-lism with magnetic resonance spectroscopy, to

diusion tensor imaging.In the second part, a comprehensive and state-

nearly 10 years has proved to be safe and reliable

as well as applicable to neurosurgical procedures,even if these procedures have to be adapted to theMRI environment to a certain extent. Neverthe-

less, the optimal solution for intraoperative imag-ing setups, combining excellent image quality withsmooth operating room work ow integration and

ergonomic comfort for the neurosurgeon, stilldoes not exist. All installed systems are prototypeswith certain drawbacks. There are dierent con-cepts with respect to scanner and operating room

design; intraoperative imaging necessitates oper-ating directly in a scanner with the drawback ofrestricted space for the surgeon or some kind of

intraoperative transport of the patient or thescanner itself. There are dierent operating tablePref

Intraoperative MR

Guest E

MRI has become a routine pre- and post-operative imaging modality in the treatment of

brain tumors and epilepsy. In the last 20 years,signicant progress in scanning technology hasresulted in high-resolution three-dimensional

anatomic imaging of the brain. In addition toanatomic imaging, information on function andmetabolism in the individual patient is available.

Since the mid-1990s, even the intraoperative ap-plication of MRI has been possible and hasopened new avenues in immediate intraoperative

Rudolf Fahlbusch, MD

Neurosurg Clin N Aof-the art overview of the intraoperative ap-

plication of MRI technology is provided. Expertsusing dierent low-, middle-, and high-eld MRIsystems available from 0.12 to 1.5 T focus on

1042-3680/05/$ - see front matter 2004 Elsevier Inc. All rdoi:10.1016/j.nec.2004.07.012ace

I Developments

ditors

dierent aspects, such as integration of naviga-tion, glioma resection, pituitary adenomas, bi-

opsies, epilepsy, and functional imaging, followedby a perspective outlook.

With the development of open MRI systems

in the mid-1990s, the concept of intraoperativeimaging, up to then only realized with CT andultrasound, experienced a renaissance. The rst

designs were based on low-eld magnets, withmagnetic eld strengths up to 0.5 T. The use ofMRI scanners in the operating environment for

Christopher Nimsky, MD

m 16 (2005) xixiiiconcepts, ranging from patient transport with an

air-cushioned operating room table to an adjacentoperating room, to movement of the patient alongthe longitudinal axis of the scanner to reach the

ights reserved.

neurosurgery.theclinics.com

FACfringe magnetic elds, to the use of some rotatingmechanism with an operating room table adaptedto the scanner. Also, the issues of MRI-compat-

ible head xation and coil design for intraoper-ative use have not yet been resolved withoutdrawbacks. Regarding the overall operating roomdesign, there are also dierent concepts, ranging

from systems dedicated for intraoperative useonly to hybrid systems combining intraoperativeuse with the application of the scanner for routine

radiologic diagnostics.To date, there is also no denite consensus as

to which direction intraoperative MRI systems

will develop. The current extremes range fromlow-eld movable installations at 0.12 T up toconcepts integrating ultrahigh-eld strength im-aging at 3 T in the operating room. Whether new

scanner designs with larger and shorter bores orthe application of dierent physical principles thatallow at MRI scanners (eg, below the operating

table) will contribute to optimizing the intraoper-ative application of MRI technology further isnot yet decided. Going to higher magnetic eld

strengths allows having a better signal-to-noiseratio, shorter scanning times, and a better resolu-tion in certain modalities, such as functional

imaging and spectroscopy. With regard to ultra-high-eld MRI, however, there may be increasedproblems with artifacts as well as geometric imagedistortions and restrictions caused by the specic

absorption rates, because the deposited radio-frequency energy must be considered in thesequence design to obtain the same performance

as in 1.5-T setups. Conversely, imaging techniquesin the direction of ultralow-eld MRI, whichcould be based on taking advantage of certain

contrast media eects relating to the Overhausereect, do not seem to be an alternative to ana-tomic patient imaging yet, even though some earlysuccess has been achieved in small animal imag-

ing. The optimal solution would be a nearlyinvisible imaging system giving online real-timefeedback to the neurosurgeon without disturbing

the surgical work ow.Meanwhile, it is agreed that intraoperative

anatomic imaging is not sucient alone. Intra-

operative imaging has to be combined with intra-operative guidance, implemented, for example, inthe form of microscope-based navigation. There

has to be the possibility to use intraoperativeimages for guidance, allowing so-called updatingof the navigation, which compensates for the eectsof brain shift. Furthermore, and of paramount

importance, is the integration of functional data,

xii PREsuch as functional MRI (fMRI) identifying elo-quent cortical brain areas and diusion tensorimaging data identifying major white matter tracts

aswell asmagnetic resonance spectroscopy for dataon metabolism. All these functional modalitiesshould also be available during surgery, reectingthe current status of the brain with respect to anat-

omy, function, and metabolism. Increasingly, de-tailed brain mapping, rendering the whole brain aseloquent, has to address the problem of infor-

mation overow for the surgeon in the operatingtheater. In addition, adequate functional para-digms have to be developed further and standard-

ized, especially with respect to their intraoperativeapplication. Even nowadays, speech mapping byfMRI is not yet standardized enough for reliablepre- and intraoperative localization. In addi-

tion to guidance maintained by navigation sys-tems, integration of robotic devices is underdevelopment.

Another important aspect of intraoperativeMRI is its acceptance in overall society. Thisseems to be no problem with regard to the patients

beneting from this technology; however, accep-tance is still ambivalent among physicians, publicopinion, and politicians as well as health insurance

providers. Intraoperative imaging per se seems tobe more and more accepted as immediate qualitycontrol during surgery. In the case of high-qualityintraoperative imaging, early follow-up imaging

(up to 3 months) is not necessary any longer.Intraoperative MRI is in competition with ultra-sound and CT as an alternative intraoperative

imaging modality, however. Recent technical de-velopments, especially in the eld of CT, allowinghigh isotropic resolution, may have the conse-

quence that these imaging technologies have to beconsidered as alternative intraoperative imagingmodalities in neurosurgery, especially if economicrestrictions are considered. Detailed economic

analyses exceeding previous preliminary cost-benet analyses must address these aspects. Pre-liminary results presented recently by Hall et al [1]

have to be extended and evaluated on a broaderplatform for industry, insurance companies, poli-ticians, and physicians. Furthermore, the signi-

cance of MRI as an intraoperative imagingmodality has to be seen in competition with otherimaging modalities, especially in operating room

setups designed for the simultaneous use by othersurgical disciplines.

In the future, perhaps as an alternative to theexpensive and highly advanced setups allowing

the identical armamentarium for pre- and

E

intraoperative diagnostics, it will be possible tohave a less cost-intensive system for intraoper-ative imaging. Such a system, based on whateverimaging modality, must generate detailed ana-

tomic information about the intraoperative situ-ation in which preoperative data on functionand metabolism have to be integrated applying

advanced mathematical techniques, includingnonlinear registration techniques as well as math-ematical simulations and models. None of these

techniques are yet robust and time-ecientenough that they can be applied for intraoper-ative use.

Intraoperative imaging is well established,especially with respect to the completion ofsurgical resections in complicated procedures;however, it is an open question as to which

direction intraoperative imaging will take. Theproblem of the practicability of intraoperativeMRI is under investigation, whether it is in the

hands of neuroradiologists and performed bythem or by neurosurgeons. Intraoperative MRIvaries from simple image generation to advanced

image processing at a high scientic level. Theexperts working on the latter level should be

obliged to present their ndings on the applicationof the method objectively.

Reference

[1] Hall WA, Kowalik K, Liu H, Truwit CL,

Kucharezyk J. Costs and benets of intraoperative

MR-guided brain tumor resection. Acta Neurochir

Suppl 2003;85:13742.

Rudolf Fahlbusch, MDDepartment of Neurosurgery

University Erlangen-Nuremberg

Schwabachanlage 691054 Erlangen, Germany

E-mail address: fahlbusch@nch.imed.

uni-erlangen.de

Christopher Nimsky, MDDepartment of Neurosurgery

University Erlangen-NurembergSchwabachanlage 6

91054 Erlangen, Germany

E-mail address: nimsky@nch.imed.uni-erlangen.de

xiiiPREFACE

nGR

magnetic dipole is then created. Not only protons with the magnetic eld creates the lowest steady-

Abut any atom that has an odd number of protons or

neutrons has a net unbalanced nuclear spin, andthus a nuclear magnetic moment. Electrons alsopossess spin and charge, and thus have a magnetic

dipole associated with them. Elements containingunpaired electrons, that is, those in which the

state energy. For example, a compass aligns its

positive pole with the South Pole of the earth,with opposites attracting. A compass can haveany orientation with respect to an external mag-

netic eld, and with it, any energy of interactionfrom zero to the maximum. Things are not quiteBasic principles of magWendell A.

Riverwoods Imaging Center, 280 West

The discovery of nuclear magnetic resonance(NMR) by Purcell et al [1] and Bloch et al [2] rstrevolutionized analytic chemistry and then medi-

cal imaging. NMR imaging has taken us to yetanother dimension of diagnostic imaging in whichsuperior contrast resolution; multiplanar capabil-ities; and imaging of physiologic processes, such as

blood ow, perfusion, diusion, cortical activa-tion, metabolite concentrations, and motion, haveprovided an entire new world of insight into the

nervous system. It is an ironic historical curiositythat the name NMR imaging was changed to MRIbecause of the publics perceived fear of nuclear

devices, because MRI uses no ionizing radiation.The fundamental interaction of atomic particles

and radiofrequency (RF) energy allows us to create

spectacularMRI scans on a routine basis. Throughrecent discoveries in physics, we know that one ofthe most fundamental particles in nature is thequark [3]. A basic property of subatomic particles

is that they possess spin and angular momentum.Within the proton, there are two quarks that spinparallel to each other and a third that spins

opposite, giving a net unopposed spin. We alsoknow that a proton has a net1 positive charge. Amoving charge produces a magnetic eld. In fact,

magnetism is dened by the force created bya specic quantity of moving charge. A tiny

Neurosurg Clin Nelectrons are not paired in outer orbitals and in

which spins are not canceled, also have an eective

E-mail address: wgibby@novarad.net

1042-3680/05/$ - see front matter 2004 Elsevier Inc. All rdoi:10.1016/j.nec.2004.08.017etic resonance imagingibby, MD

iverpark Drive, Provo, UT 84604, USA

magnetic moment. The magnetic moment associ-ated with an electron is approximately 1000 timesgreater than that of a proton.

In this article, no attempt is made to denerigorously with mathematic techniques the inter-actions of the nuclei with each other and withexternal energy. Rather, an attempt to explain

these concepts through the use of simple physicalmodels that speak a universal language is made. Ofcourse, no physical model is able to explain the

nature of subatomic particles completely, just asno single mathematic equation currently explainsthe dual nature of matter. A number of earlier

articles on the basics of MRI [410] are includedwithin the references for the interested reader.I recognize that this article may go into far more

detail than the typical reader requires. Neverthe-less, for those few brave, intrepid, and curioussouls who really wish to know what is going on inthe mysterious insides of an MRI scanner, I have

tried to make this model as complete as possible.Having a basic understanding of these principlesallows one to optimize image quality, reduce error,

and improve conspicuity of pathologic ndings.A cookbook approach gives mediocrity at best.

We are all familiar with the property of

a magnet, which when placed within a magneticeld, aligns itself in such a way that its interaction

m 16 (2005) 164as simple at the atomic level. By quantum theory,

only certain energy states are allowed, which arediscrete in value. The hydrogen nucleus havinga spin quantum number of positive 12 and negative12 gives dipole vectors that point 35.26( with and

ights reserved.

neurosurgery.theclinics.com

te

t

r

a

2 GIBBYagainst the magnetic eld [5] as illustrated inFig. 1.

Precession

When rst placed in a magnetic eld, the o-axis proton dipoles begin to precess at a rateknown as the Larmour frequency. The often-used

analogy of a spinning top precessing under theforce of the earths gravitational eld is illustratedin Fig. 2. An important point to remember is that

the motion of the precessing magnetic dipole andthe motion of the atom are completely indepen-dent. The small spinning dipole within the nucleus

maintains its orientation relative to the magneticeld in spite of rapid molecular tumbling andtranslational motion caused by thermal energy

within the lattice of the molecular structures.There are only two things that inuence the

precessional rate (angular velocity) of the spinningdipole. Each dierent element, be it a single pro-

ton, a nucleus composed of many protons andneutrons, or an electron, has a dierent angularmomentum, and thus a dierent precessional rate

Fig. 1. Hydrogen nucleus with spin quantum numbers

of positive 12 and negative12 . The magnetic dipoles

reside in energy states pointing with and against the

magnetic eld. The vectors pointing against the magnetic

eld are in a higher energy state.

Fig. 3. (A) The vectors oriented with B0 spin in an opposi

Most of these cancel each other out. We are left with net vec

eld (approximately 1 of 100,000 vectors). (B) The nuclea

lattice. There is only a tiny energy dierence between the up

lowest energy state (pointing up).for a given magnetic eld strength [5]. The pre-cessional frequency of hydrogen in a magnetic

eld at 1.5 T is 63.866 MHz, whereas that forphosphorous is 25.876 MHz [11].

The second critical element in determining how

fast a proton precesses is the net magnetic eldthat it experiences. A top that is precessing on themoon precesses at a dierent velocity than if itwere spinning on the earth because of a dierence

in the gravitational force. Likewise, protonsspinning in dierent magnetic eld strengths pre-cess at dierent velocities. This is given by the

relation: Frequency = c b, where c equals thegyromagnetic ratio (for hydrogen, c= 2.6751978 108s1T1) and b equals the eld strength intesla [12]. Our model now describes an ensembleof precessing magnetic dipoles with their vectorspointed in the direction of the magnetic eld.

Fig. 2. A spinning top oriented o-axis with earths

gravitational eld experiences two forces: the gravita-

tional eld, G, tending to pull the top toward the earth

and an opposing centrifugal force, F, from the spin of

the top. The result is a wobbling or precessional motion

around earths gravitational eld. A similar motion

occurs with spinning magnetic dipoles when placed

within a magnetic eld.

direction than those oriented opposite the magnetic eld.

ors precessing around the z-axis oriented with the magnetic

dipoles exchange energy with the surrounding molecular

nd down states, leaving a small net fraction of dipoles in the

c

3BASIC PRINCIPLES OF MAGNETIC IMAGING

4 GIBThermal motion

The second factor that causes importantchanges in our model is the eect of thermal

motion. The energy dierence between the twostates (ie, for or against the magnetic eld) is smallcompared with other energy transitions. Forexample, a typical x-ray produced by the de-

celeration of an electron into a tungsten anode ison the order of 100,000 electron volts. The energyfrom the transition of electrons from outer

orbitals to inner orbitals, producing light, is onthe order of 4 electron volts. The energy transitionfor a small proton dipole pointing for or against

a magnetic eld at 1.5 T is only 2.6 107electron volts, a trillion-fold dierence from x-rays. This corresponds to energy in the RF range.Energies in this range are ubiquitous in the

environment because of thermal motion. Thelow energy of these transitions accounts for thesafety of MRI. It also requires that the detection

apparatus be extremely sensitive, however. It isa system that is noise limited.

Thus, each of the little dipoles in our model is

inuenced by the rapid exchange of thermalenergy with the surrounding molecular lattice.As the dipole absorbs energy, it is raised to an

excited state. As the relaxation process occurs,this energy is exchanged with the environment andthe dipole is aligned in a lower energy (ground)state. Because of the thermal energy available, the

proton dipoles undergo rapid shifts betweenorientations with and against the magnetic eld.If the energy of the lattice were not available,

relaxation would be extremely slow. At a giventime, only a tiny net fraction is oriented with themagnetic eld [12]. The distribution at room

temperature is given by the Boltzmann equation:

NNo

eDEkTwhich with a Taylor0s expansion; simplifies to

NN 1

DEkT

orNN 1

hcB2pkT

where k = Boltzmans constant (1.38066 1023J K1); T = Absolute temperature; Bo = Fieldstrength in Tesla; h = Plancks constant (6.062608 1034) J sec;x= cBo in rad sec1; m cBo2pwherec x2p in hz; E hm; DE hm hcBo hcBo2p ;c= gyromagnetic ratio (2.6751978 108 s1T1) for hydrogen.

This is eld (Bo) dependent; thus, at 1.5 T, 9.88

of 1,000,000 are oriented with Bo, giving a largerfraction of nuclei available for excitation than at1.0 T, which has only 6.59 of 1,000,000 orientedwith Bo. This is why image quality is better athigher eld strengths; more protons are available

for excitation.Because the vectors that are oriented against

the magnetic eld spin in an opposite direction,they cancel out any vectors precessing with the

magnetic eld. For the purposes of our model, weonly need to consider the dierence between thetwo, that is, the small fraction of nuclei that

represent the dierence between the two popula-tions (Fig. 3A). We now have protons precessingat a specic frequency in a magnetic eld, which

are undergoing rapid up and down transitions,leaving only a tiny net magnetic vector, as isillustrated in Fig. 3B.

Phase coherence

The next element of our model is that of phase.Phase is a measure of the relative position of anobject or vector; usually measured with an angle /over a given time for a periodic function. A simpleanalogy may help to clarify this point. If we taketwo wheels, place a dot on one edge of them, as

shown in Fig. 4, and spin them both with the samevelocity, they remain spinning synchronously overtime. If one wheel spins slightly faster than the

other, however, the dots no longer align over time.They are said to be out of phase with eachother. A similar process happens with nuclearspins.

Although the small dipoles are placed ina strong homogeneous magnet, the magneticeld that they sense is slightly dierent than that

of their neighbors. This can be a result of severalfactors:

1. Although the magnetic eld is homogeneousby technical criteria, it may still contain an

inhomogeneity of approximately 1 part permillion (ppm) [7]. This translates into a fre-quency dierence of 63 Hz at 1.5 T. A dipolesensing a 1.5-T magnetic eld spins slightly

dierently than an adjacent proton thatexperiences a magnetic eld of 1.500001 T.In 8 milliseconds, protons in the same voxel

sensing a dierence of 1 ppm magnetic eldinhomogeneity are 180( out of phase andcancel out each others signal.

2. There are many local disturbances to themagnetic eld ranging from the molecularlevel up through the tissue level. For instance,

BY

, t

to

ti

Moxygen atoms, by their nature, are highlyelectronegative and tend to attract more ofthe shared electron cloud around themselves

than does the adjacent hydrogen atom ina molecule like water. This causes thehydrogen nucleus to have less of a screening

eect from the overlying electron cloud. Itexperiences a stronger local magnetic eldthan a proton attached to a fat molecule in

that the protons are more shielded by valenceelectrons. Therefore, the spins of waterhydrogen nuclei precess at a slightly greater

frequency than those of fat.3. Dierent substances have dierent perme-

abilities to the magnetic ux and can thusdistort magnetic eld lines. Most materials

are diamagnetic, meaning that the magneticux lines would rather go through a perfectvacuum than through that substance. For

example, at an air-bone or bonesoft tissueinterface, the magnetic ux lines are dis-torted. Some materials, such as iron, are

ferromagnetic and concentrate magnetic uxlines. All these make the local magnetic elddierent for adjacent nuclei, causing them toprecess at slightly dierent rates. Because

they rapidly become out of phase with respectto each other, and coherence is lost.

Our model now consists of small nuclearmagnetic dipoles rapidly precessing in space, each

with a slightly dierent angular velocity, out ofphase with respect to adjacent dipoles, and rapidlyexchanging energy with the environment such that

their dipole orientations are ipping back andforth in restricted quanta of energy for andagainst the magnetic eld. One may then wonder

how any useful information can be extracted fromthese weak signals, precessing at dierent veloci-ties and completely out of phase with each other.

Net vector

At this point, a short diversion is necessary to

understand signal generation. Each of the individ-ual spinning nuclear dipoles can have only one oftwo orientations with respect to the magnetic eld.

If we average the orientations of these, we canobtain a net vector that can have any orientationwith respect to themagnetic eld. This averaging of

the individual directions is illustrated in Fig. 5.This net vector can be thought of as a larger singleFig. 4. A dot is placed on a spinning wheel. In the rst case

the dots stay synchronized with one another. They are said

rst wheel is dierent from that of the second wheel. Over

BASIC PRINCIPLES OFthey precess at slightly dierent frequencies,he angular velocity of the two spinning wheels is equal and

be in phase. In the second instance, the velocity of the

me, the relative position of the dots drifts out of phase.

5AGNETIC IMAGINGdipole. For example, if sucient RF energy is given

to a sample of spinning hydrogen nuclei, a fraction interact with the oscillating magnetic eld of the

o

6 GIBBYof them can be rendered into the excited state andthe net vector tipped from 0( through perhaps 90(.This would reect a 90( RF pulse. By convention,the axes used in MRI are x, y, and z. The z-coordinate is taken along the main magnetic eld,and the protons process around z in the x-y plane.We remember that if a magnetic dipole spins, an

electrical current can be induced in a coil orientedperpendicular to this, just as the converse is seenwith a coil-conducting current that induces a mag-

netic eld. This is shown diagrammatically inFig. 6.

Signal formation

The key to obtaining any useful information

from the precessing protons is to establish co-herence, that is, having a large portion of thedipoles all spinning together with a net vector

precessing in the x-y plane, where signal can begenerated in a suitably oriented coil. How then isphase coherence established and signal generated?

RF waves are electromagnetic radiation that havetime-varying magnetic elds propagated through

spinning hydrogen dipole.To understand how phase coherence is

achieved, we must rst understand the B1 eld

created by the transmitting RF coil. In Fig. 7,a loop of wire is fed an oscillating RF current. Aswe remember from our basic college physics,current owing through a wire induces a magnetic

eld perpendicular to the ow of current accord-ing to the right hand rule. If the current isoscillating, as occurs with a RF pulse, a sinusoidal

oscillation of the B1-induced magnetic eld of the

Fig. 6. An electromagnetic pickup coil oriented perpen-

dicular to the spinning dipole acts as a tiny generator inFig. 5. The concept of a net vector. Each of the individual n

for or against the magnetic eld, the net vector is the sumspace. These time-varying magnetic elds canuclear dipoles is added. Now, rather than pointing at 35.3(f the positions of all the vectors.which an oscillating electrical current is created.

time along the axis of the spokes (Fig. 8). Most perpendicular to the axis of a y wheel. The B1

cr

7MAGNETIC IMAGINGRF transmitters use a circularly polarized trans-mitting coil, which creates a wave of an oscillatingB1 magnetic eld that rotates around the sample

with a B1 vector going in and out perpendicular tothe z-axis (Fig. 9).

The next thing to understand is spin locking. If

we looked at only the vectors in the x-y planefrom our net vector in Fig. 3B, it would look likeFig. 10. The vectors are precessing in the x-y planeall out of phase with respect to each other.

Spin locking or synchronization of the vectorsoccurs as the spin dipoles are pushed togetherby the synchronized B1 eld of the RF transmit

coil. Some authors use a rotating frame ofreference, a mathematic trick in which the ob-server rotates around the vectors to describe

a classic model of the interaction with thespinning vectors. I prefer to view this from a morelinear approach. In Fig. 11, the circular motion of

a given vector of the spinning nucleus observedfrom one point of the transmit coil can be viewedas a series of vectors of increasing and then

vector acts like a piston opposing or reinforcingthe motion of the spinning hydrogen vector(Fig. 12). The B1 vector created by the oscillating

magnetic eld of the RF wave quickly brings thevector of the nucleus into synchronization; how-ever, this does not occur instantly. The time

constant for this process is called T1rho. If wethink of an ensemble of nuclei spinning in a voxelin dierent phases, the B1 vector moves coherent-ly in a circularly polarized fashion around the

spinning nucleus, with all the B1 vectors in phasefrom the oscillating RF transmit coil. This quicklysynchronizes all the opposing vectors, bringing

them into coherence much like independent pis-tons working with or against a large coherentlyoperated crank shaft with synchronized pistons

(Fig. 13).Thus, with the tiny individual magnetic dipoles

in phase, we can add each of them together, giving

a large single magnetic dipole. With coherencenow established and a net dipole vector precessingin the x-y plane, a signal is generated in the coil orcoil is created. The transmitting coil createsa magnetic vector perpendicular to its face, whichincreases and decreases and then reverses overtime as a function of the RF oscillation. If one

thinks back to our spinning wheel analogy, themagnetic vector would be going in and out over

Fig. 7. Passing an oscillating current through a coiled wire

coil that moves in and out perpendicular to the face of the

BASIC PRINCIPLES OFdecreasing sinusoidal intensity over time. In otherwords, the vector pointing at the coil is doingexactly the same thing that the oscillating trans-mitted RF wave is doing by creating a vector thatgoes in and out perpendicular to the z-axis. Think

of the component of the vector facing the coil assomething like a piston that oscillates in and out

eates an oscillating magnetic eld (B1) perpendicular to the

coil as a function of time.antenna. This signal represents a free induction

of

8 GIBBYoccurred within the system. First, a certain frac-

tion of the nuclei were inverted into the excitedstate against the main magnetic eld, giving usa net magnetic vector that is not oriented with the

longitudinal z-coordinate. Second, phase coher-ence that had not previously been present was

eld strength. As more of the protons relax,however, a smaller percentage of nuclei are in

the excited state; thus, fewer are available overallto relax. This describes a typical exponentialrelaxation curve, as illustrated in Fig. 15. At

a time of 1 T1, 64% of the longitudinal

Fig. 9. A circularly polarized transmit coil creates an oscillating B1 eld that moves in and out perpendicular to the z-

axis that rotates around the sample.decay (FID). Of course, this does not last long.Because the precession rate for the nuclear dipolesvaries with magnetic eld dierences, phase co-

herence is lost and the signal rapidly decays, asdemonstrated in Fig. 14. The length of time thesignal persists is a measure of how rapidly phase

coherence is lost.By applying RF energy, a two-part change has

Fig. 8. The transmitting coil creates an oscillating B1 eld

of the turning wheel, which would be the representation oestablished. Over time, the system returns to itsnatural random state.

T1 relaxation

The rate at which the proton dipoles relax back

into the aligned state with Bo (lowest energy) isconstant over time for a given substance at a given

ver time that moves in and out perpendicular to the spokes

the precessing vectors in the x-y plane.

9BASIC PRINCIPLES OF MAGNETIC IMAGINGmagnetization has recovered. By four T1, 98% ofthe longitudinal magnetization vector has beenrecovered. The process of changing from an ex-cited to a nonexcited orientation in the magnetic

eld involves an energy exchange, the energy ofwhich is precisely equal to the Larmour frequencyof the spinning nuclear vector. What MRI is really

seeing inside the body is energythe energy ofprecessing nuclear dipoles.

As previously noted, these energies of ex-

change can be obtained from molecular motionsand vibrations in the molecular lattice. For thisreason, T1 is commonly referred to as the spin

lattice relaxation time. From quantum theory,only a discrete energy value is allowed to inducethis transition. Energy is related to frequency, t,by the equation E = hm, where h is equal toPlancks constant [13]. An important concept isthat the excited magnetic dipole can relax only if itcan transfer its discrete energy into the surround-

ing molecular lattice. These molecular energystates are present in the form of rotational andvibrational motions of the molecules. Certain

types and structures of molecules are far moreecient in accepting these energies, because theirvibrational and rotational energies correspondmore closely with the Larmour frequency. Fre-

quencies that are too high or too low do noteciently interact with the nuclear dipole; thus,T1 relaxation is slowed [14].

For example, water molecules are small. Theyrotate and vibrate quickly and have a relatively

Fig. 10. Top down view of Fig. 3 demonstrates the

precessing x-y component of the vectors spinning

around randomly completely out of phase with respect

to each other.

Fig. 11. If one stands at a given vantage point watching

the precessing nuclear dipole, the vector increases and

then decreases in a sinusoidal oscillating function over

time.

Fig. 12. The oscillating B1 vector exerts a force on the

spinning proton nuclear vector much like two opposing

pistons in a cylinder.

BYhigher spectral frequency of lattice energies. Purewater contains little of the spectral energy needed

to induce T1 relaxation of the small nucleardipoles. Conversely, fat molecules that tumblemore slowly have a spectral energy more closely

matched to the Larmour frequency and henceallow for more ecient relaxation. When theprotons undergo faster T1 relaxation, more of

their longitudinal vector is available for eachsucceeding pulse. Therefore, more signal is gen-erated, because a larger vector is available toprecess in the x-y plane. That substance appears

relatively brighter. For this reason, fat is brighton T1-weighted (T1W) images (images thataccentuate dierences in the T1 of tissues) and

water is dark (Fig. 16). Likewise, myelin, whichhas a slowing eect on the motion of adjacent

water, is relatively bright on T1W sequences [15].Nevertheless, this can be pushed too far. Ex-

tremely large solid-like structures, such as boneor proteins (eg, ligaments and other highlyordered proteins), have protons that are relatively

immobile. They give little signal on T1W imag-ing, because the rotational and vibrational fre-quencies have been slowed to the point that they

are no longer optimal for relaxation [12,16].Similarly, protons on cholesterol and lipid mem-branes have relatively poor mobility and havelonger relaxation times as opposed to adipose

tissue (storage fat), which has molecules that arein an oil (liquid) state, are more mobile, and relaxmore quickly. Paramagnetic materials also im-

prove T1 relaxation, as can nonparamagneticcalcium salts [17].

Fig. 13. (A) An assembly of pistons on the bottom row represents each of the dierent proton vectors at dierent phases,

creating a net vector of zero. (B) With the application of the synchronized force of B1 represented by an array of pistons

at the top of the diagram, the vectors of the precessing protons are rapidly brought into synchronization, giving a strong

net vector.10 GIB

ms

ca

lfst

vl

ioa

rlm

mtm

11BASIC PRINCIPLES OF MAGNETIC IMAGINGBy the same token, the relaxation of biologic

aterial is more or less ecient depending on thetrength of the magnetic eld. T1 relaxation iseld strength dependent [18,19]. Table 1 gives data

omparing T1 relaxation times of selected tissuest 24 and 2.5 MHz [105].In general, T1 relaxation is more ecient for

ower frequency (eld strength is proportional torequency) over the range of magnetic eldtrengths used clinically. Thus, shorter repetitionimes (TRs) can be used for a 0.35-T magnet

ersus a 1.5-T magnet to achieve equal T1 re-axation (and hence T1 contrast between tissues).T1 relaxation is a thermodynamic process

nvolving enthalpy, in that an energy exchangeccurs. The surrounding lattice must be able toccept the precise quanta of energy emitted by the

elaxing nuclear dipoles. The dierence in re-axation between small, intermediate and largeolecules is illustrated in Fig. 17. T1 can be

easured by sampling the system at variousime intervals to see how much longitudinal

time. To measure T1 accurately, the TR (ie, the

time before the system is re-excited) must spanvalues above and below T1. The TR is varied, andsignal intensity is plotted as a function of TR.

The slope of the curve is related to T1: Signal =Mo (1 eTR/T1) [20].

Fig. 13 (continued )

Fig. 14. When all the vectors are spinning in synchrony,

the maximum signal intensity is generated in the coil.

Over time, however, these vector dephase, and there is

rapid loss of signal intensity. This represents freeagnetization is present after a given amount of induction decay (FID).

to

a

12 GIBBYtion of the local magnetic eld by materials thathave dierent magnetic permeability, such asbone, air, or stationary paramagnetic or ferro-

magnetic materials. Because these magnetic eld

Fig. 16. Orbital MRI scan with T1 weighting. Fat is

bright, bone is dark, muscles are low signal, and vitreous

humor is dark. Note that the lens is slightly brighter than

the vitreous uid, because the water is bound to

proteins, which slows its motion.

Table 1

T1 relaxation values

24 MHz

(milliseconds)

2.5 MHz

(milliseconds)

Clotted white

blood

867 404

Serum 1590 820

Gray matter 644 332

White matter 469 264

From Ling CR, Foster MA, Hutchison JMS. Com-

parison of NMR water proton T1 relaxation times of

rabbit tissues at 24 MHz and 2.5 MHz. Phys Med Biol

1980;25:748; with permission.T2 relaxation

The other decay process that is occurringsimultaneously with T1 relaxation is the loss of

phase coherence. This is termed T2 relaxation orspin-spin dephasing (ie, one spin becomes out ofphase with another spin). As opposed to T1

relaxation, the loss of phase coherence is not onethat requires an exchange of energy. In chemicalterms, it is a process involving entropy, or a dis-

ordering of an ordered state. This also occurs atan exponential rate [20] (there is more signal todephase early on): Signal =Mo e

(TE/T2). As thespins dephase, the magnetic vectors precessing inthe x-y plane gradually fan out. After a certain

63

86939899

1T1 2T1

Magnetiza

% R

ecov

ery

Long

itudi

nal M

agne

tizat

ion

Fig. 15. Graphic representation of T1 relaxation. The exp

63% of the signal intensity has recovered. By 5T1, 99% hperiod, all phase coherence is lost and no moresignal is generated. At a time of T2, 64% of thephase coherence is lost.

The loss of phase coherence is caused by onlyone thinga slightly dierent magnetic eldexperienced by adjacent spinning proton dipoles.

This slightly dierent magnetic eld can beachieved by many dierent processes, however.These dierences in magnetic eld can be sub-

divided into two major categories.

Static magnetic elds

Static magnetic elds vary in intensity overspace but not over time during image acquisition.Examples of this would include magnetic eld

inhomogeneities by the magnet itself, perturba-

3T1 4T1 5T1Time

ion = Mz(1- e-TR/T1)nential recovery of T1 demonstrates that at a time of 1T1,

s recovered.

13BASIC PRINCIPLES OF MAGNETIC IMAGINGinhomogeneities are constant in time, the signalloss from dephasing can be recovered by the useof a second 180( pulse, which rephases thenuclear spins (more about this in the section on

Time-varying magnetic elds

Water molecules, for instance, can move rap-idly through space and across membranes and canrandomly bounce around within and between

Fig. 17. (A) Water molecules tumble rapidly and have a large population of high frequency vibrational and tumbling

energy states. As a result, T1 relaxation is inecient. (B) Fat molecules, on the other hand have a larger proportion of

motional states correlating with the energy needed for relaxation. (C) Complex molecules such as membranes or water

bound to large protein molecules exhibit very slow motion resulting in low frequency components that are below the

energy needed for relaxation. (D) A composite gure demonstrates that fat will have better eciency at relaxation than

eigher water or solid materials. This is the result of the quantum requirement for discrete energy transitions which can be

supplied only with certain molecular vibrational and rotational states.pulse sequences).

14 GIBBYadjacent voxels. The water molecule that movesinto a new area of dierent magnetic eld strength

precesses at a slightly dierent rate and thereforebecomes out of phase. Because it is not static,however, a reversal of its spin cannot bring it backinto phase coherence, and this part of the signal is

lost. T2* is the combined loss of phase coherencefrom static and time-varying magnetic eld in-homogeneity. The dierence between phase lost

from static and time-varying magnetic eld in-homogeneities is shown diagrammatically inFig. 18.

Phase loss can occur not only between adjacentvoxels or imaging points in our data set but withina voxel as well. For example, fat and water precess

at slightly dierent frequencies. The hydrogen ofa water molecule, being less shielded by theelectron cloud of its oxygen neighbor, experiencesa higher magnetic eld strength and thus precesses

at a faster rate. If a voxel contains equal quantitiesof fat and water, they are exactly out of phasewith each other at certain times causing the signal

to cancel.

Fig. 18. T2 relaxation versus T2* relaxation. Phase loss

occurs in two ways: reversible phase loss and irreversible

phase loss. T2* is a combination of the phase losses of

static (reversible) and nonstatic (irreversible) eld in-

homogeneities. Using a spin echo pulse technique, the

signal loss induced from static magnetic eld inhomo-

geneities can be recovered and the T2 relaxation

measured. This gure graphically illustrates why the

T2-weighted sequences always have greater signal in-

tensity than a T2*-weighted sequence for a constant

echo time.T2 relaxation is greatly augmented by havinga distorted magnetic eld. Dierent tissues have

dierent T2 relaxation rates depending on theirphysiochemical constituents. Water tumbles rap-idly in space. As it does, any magnetic eld

distortions are rapidly averaged out over time.Thus, adjacent water molecules all experiencea similar magnetic eld, and their nuclear dipoles

dephase slowly. Let us suppose, however, thata protein or large polysaccharide molecule isintroduced into the solution. The water moleculesbound to the biologic polymer rotate more slowly

than adjacent free water, and magnetic eldinhomogeneities are averaged less well. Becauseof this, the water molecules in dierent hydration

states experience dierent magnetic eld strengthsover the period of image acquisition. An excellentexample of this is the lens of the eye. The rigidly

held water molecules in proteins rapidly losephase coherence, generate a small signal for imag-ing, and thus are dark on T2-weighted (T2W)images. Fat, being a much larger molecule than

water, is held more rigidly in space over time andthus loses phase coherence more rapidly than freeextracellular water; fat darkens relative to water

with increased T2 weighting (Fig. 19).If water molecules are adjacent to or within

voxels containing materials that cause distortions

of the magnetic eld, they likewise rapidly losephase coherence. For example, iron deposits inthe basal ganglia destroy phase coherence, giving

little signal on T2W images. Likewise, the in-jection of magnetite, a ferromagnetic substancethat causes strong local eld inhomogeneities,

Fig. 19. Axial T2-weighted orbit image. Fat is darker,

the lens is dark (complex protein), and the vitreous

(water) is bright.

energy pulse at this point in time, the entire vector

a

a

r

15BASIC PRINCIPLES OF MAGNETIC IMAGINGof the short T1 substance labeled A can be rotatedinto the x-y plane and is available for generatingsignal. Those substances with intermediate andlonger T1s (B and C, respectively) have less

magnetization available to precess in the x-y planeand generate less signal in our receiver. Thus,maximum signal is obtained by waiting a longer

period for full longitudinal magnetization recov-ery to occur or by speeding the T1 relaxation ofthe slower substances through the use of para-

magnetic agents. Dierential intensity betweenvoxels of dierent T1s can be achieved by select-ing a TR close to the T1 of the tissue of interest.

To repeat, the brightest tissues on T1W pulsesequences are those that have the shortest T1 andthus have the most available longitudinal magne-tization available for inversion into the x-y plane

Rigid molecules

Bone

Fibrocartilage

Ligaments Dark Dark

Scar

Hemosiderin

Watery substances

Cerebrospinal uid Dark Bright

Cysts

Free water

Intermediate molecules

Fat Bright Intermediate

Proteinaceous

material

Intermediate

to bright

Intermediate to

bright depending

on water content

Hyaline

cartilage

Intermediate Intermediate to

dark

Lens of eye Bright Darkdestroys phase coherence and gives dark signalon T2W images. Materials that are more solidmove more slowly and thus have a more in-

homogeneous local magnetic eld. Fig. 20, illus-trates how a semisolid material distorts the localmagnetic eld compared with a rapidly tumbling

small molecule. For most substances, phase islost much more quickly than restoration ofproton dipole alignment with the z-axis (ie, T2relaxation is much shorter than T1 relaxation). A

comparison of various substances is given ina qualitative way in Table 2.

Fig. 20. Comparison of a small versus large molecules on m

water molecules evens out micro magnetic distortions giving

molecules do not tumble as rapidly. Therefore, small pertu

Table 2

T1-weighted and T2-weighted appearance of various

body tissues

Appearance on sequences

T1-weighted T2-weightedRecovery of magnetization vector to ground state

(T1 relaxation)

Let us return to our model briey. After an RF

pulse is given, a magnetization vector is estab-lished that is precessing coherently in the x-yplane. Suppose that a sample contains a variety of

substances that have dierent T1 relaxation values(A, B, and C for short, intermediate, and longT1s, respectively). Given a long enough time, allthe magnetization vectors of the various substan-

ces return to an equilibrium position along the z-axis. If we excite the system at an intermediatetime, those voxels with short T1s will have already

relaxed, with their magnetization vectors orientedparallel to the z-axis before the next excitation.Tissues with longer T1s will be somewhere in

between. Their relative positions are shown inFig. 21. If the sample is given a repeat 90( RF

gnetic eld homogeneity. (A) The rapid tumbling motion of

homogeneous magnetic eld. (B) Large solid or semi-solid

bations of the local magnetic led occur.

16 GIBBYat the beginning of each pulse repetition. Table 3gives the values of T1 for various tissues. Re-

member, T1 relaxation is eld strength dependent;therefore, it is not possible to use values obtainedat high eld strength to compare with low eld

measurements and vice versa. One should alsoknow that there is considerable variability in themeasurement of T1 between dierent instruments

and dierent investigators. As such, the T1 re-laxation of a tissue is not useful as an absolutecomparison with other disease processes. Fieldstrength, equipment type, temperature, selection

of the sampling sequence, TR [21], and slice

How fast tissues lose coherence

Paradoxically, tissues that have a short T2 have

the least signal. This is easily understood by thefact that as phase is lost, signal is destroyed andimage intensity is decreased. Substances like free

water, whichmaintain phase coherence for a longerperiod, have relatively greater signal than tissueswith a short T2 if the data sampling (related to

echo time [TE]) is taken at longer and longer timesafter excitation. Almost all pathologic processes(eg, tumors, inammatory disease, infections,trauma) result in increased water content and

edema. For this reason, they are best seen on

Fig. 21. Comparison of T1 relaxation between substances with short, intermediate, and long T1s. (A) In the rst

instance, vectors A, B, and C are excited into the x-y plane. A long time follows before the pulse is repeated. At such

time, TR, repitition time, all the vectors will be back to their ground state and ready for full excitation into the x-y plane.

(B) A similar experiment is performed, except that a short TR is used. At the time that a new 90( radiofrequency (RF)pulse is delivered, substance A with a short T1 will be tipped fully into the x-y plane, giving the largest signal. Its net

vector is larger than that of a substance such as C, which has a long T1 relaxation time that has not fully relaxed before

excitation, yielding a smaller net vector in the x-y plane.thickness all inuence the obtained value.

se

m

it

ce

17BASIC PRINCIPLES OF MAGNETIC IMAGINGFig. 22. A patient studied with Carr-Purcell-Meiboom-Gill

81, and 160 milliseconds. (A) At the lower echo time (TE; 31

Notice that the white matter is relatively dark compared w

milliseconds (C). With an extremely long TE, however, the

of the gray/white dierentiation.quence, multiecho, T2-weighted images with echoes at 31,

illiseconds), the best gray/white dierentiation is achieved.

h the gray matter at a TE of 81 milliseconds (B) and 160

rebrospinal uid is prominently displayed, but there is lossT2W images. The optimum TE should be thatclosest to the T2 of the tissue of interest [22]. For

instance, a cyst or area of edema may have long T2times; sampling is best done at long TEs todistinguish the aected area from adjacent tumor

tissues. If one wishes to distinguish between grayand white matter (Fig. 22), such long TEs are notadvantageous, because a signicant amount of

signal is lost; the tissue is sampled long after theoptimum dierence between the signal intensities

is observed. If one wishes to view only cerebrospi-nal uid (CSF), a long T2 of 200 millisecondsprovides a myelographic eect (Fig. 23). Suchrelative signal intensity for dierent substances as

a function of TE is illustrated in Fig. 24. Table 4gives T2 relaxation values for various biologictissues and uids. T2 relaxation, unlike T1 re-

laxation, is not eld dependent. Equipment var-iances and dierences in sampling techniques haveled to wide variations in reported T2 values of

dierent tissues. Again, these cannot be used tocompare absolute values. However, on a givenMRI machine, the reproducibility of T1 and T2

measurements is excellent, ranging from 5% to 9%variance [23].

Initially, it was hoped that dierent pathologicprocesses could be dierentiated on the basis of

characteristics T1 and T2 signatures [24,25]. Un-fortunately, there is a wide overlap betweenbenign and malignant processes [2628], yielding

Table 3

T1 Relaxation values for various tissues

Brain 1.5 T 4.0 T

Gray matter 850 [21]1023 [23] 1724 [77]

White matter 550 [21]710 [23] 1043 [77]

Cerebrospinal uid 3200 [21] 4550 [77]

Fat (adipose) 200 [106]

Muscle 800 [106]

3s

m

llis

c

18 GIBBYpathologic processes is complex by relaxationmeasurements and changes as the pathologic

process develops [29].

Inte

nsity

82 133 158 186 22

Fig. 24. Hypothetic T2 decay curves for various biologic

tumors. Notice that the best time to sample the data (echo ti

to distinguish between gray and white matter, a TE of 65 mi

matter and cerebrospinal uid, a TE of 145 milliseconds issignal than those with fewer hydrogen atoms (eg,water versus bone). An obvious fact is that if thehydrogen atoms move out of the plane of interest

White MatterGray Matter

CSF

BloodFatTumor

Time msec.

ubstances, such as white matter, gray matter, cysts, and

e [TE]) depends on what one is looking for [ie, if one wishes

econds is chosen; if one wishes to distinguish between gray

hosen]).little benet to measuring T1 or T2 of a given

disease. Furthermore, the characterization of

How much hydrogen is available to image

A third important parameter of signal intensityis that of the hydrogen spin density. Substancesthat contain more hydrogen atoms have more

Fig. 23. Fast spin echo (FSE), heavily T2-weighted, cor-

onal, lumbar MRI scan demonstrating excellent contrast

between the cerebrospinal uid of the subarachnoid space

and the conus medullaris (arrows). All other structures

are relatively dark. (FSE repetition time = 8000 milli-

seconds, echo train length = 16 milliseconds, echo

time = 192 milliseconds, 24-cm eld of view, 4-mm slice,

512 384 matrix, number of excitations = 2).

Table 4

T2 Relaxation values for various biologic tissues and

uids

T2

Frequency

(MHz)

White matter 65 [106]75 [23] 60

Grey matter 105 [106]85 [23] 60

Cerebrospinal uid 2000 [107] 25 [107]

Blood 250 20

Fat 200 60

Muscle 63 63

Data from dierent experiments and under dierent

conditions.

From Bottomley PA, Foster TH, Argersinger, Pfeifer

LM. A review of normal tissue hydrogen NMR relaxa-

tion times and relaxation mechanisms from 1100 MHz:

dependence on tissue type, NMR frequency, tempera-

ture, species, excision and age. Med Physic 1984;11(4):

42548; with permission.

called 90( pulse. The vector then precesses in

19BASIC PRINCIPLES OF MAGNETIC IMAGINGduring signal acquisition, signal is again lost. Thiscan be illustrated by the ow void seen withowing vessels.

We can combine these concepts into a single

equation that should not be dicult to understand[22]. The signal intensity (I) for a given sample isrelated to the number of hydrogen atoms (ie,

hydrogen spin density) given by S and to twoexponential decay components: the T1 relaxation

given by 1 eTRT1 and the T2 relaxation given bye

TET2; therefore, I S 1 eTRT1 [eTET2]. For

signal to be acquired, the magnetization vectormust precess within the x-y plane and must be

coherent.To summarize our model thus far, the spinning

of small subatomic particles creates a small mag-

netic dipole. The energy of interaction withan externally applied magnetic eld can exist onlyin discrete energy states (quantum levels). Forthe hydrogen proton, there are two allowed. In

a magnetic eld, these nuclear dipoles precessaround the z-axis. A thermal equilibrium isestablished with populations of dipoles spinning

with and against the applied magnetic eld. Whensuitable energy is given in the form of RF, a smallfraction of these hydrogen nuclei can be excited

and brought into phase coherence, precessing inthe x-y plane. Over time, their orientations re-establish equilibrium with the magnetic eld (Bo)

at a rate given by the exponential time constantT1. They lose phase coherence with an exponen-tial time constant known as T2. Their precessionalrate is only dependent on the net local magnetic

eld experienced by the nucleus. T1 relaxation isthus governed by how quickly the nuclei exchangeenergy with the lattice or surrounding molecules,

and T2 relaxation is governed by local magneticeld inhomogeneities.

Basic pulse sequences

I have purposely been vague up to this point asto the exact pulse sequences and gradients that areneeded to establish this. So far, we have only given

the example of a 90( RF pulse causing an FID.A bewildering array of pulse sequences is

available for MRI [30]. Slight variations on these

sequences have led to various acronyms. Somepulse sequences are nearly synonymous with oridentical to others but have been given dierent

names by dierent authors. Spin echo (SE),inversion recovery (IR), short time inversion re-covery (STIR), gradient-recalled acquisition in thesteady state (GRASS), steady-state free preces-sion (SSFP), Carr-Purcell-Meiboom-Gill (CPMG)sequence, to name only a few, are included in thecurrent literature. On top of that, with each new

pulse sequence modied by variations of gradientsand acquisition times, equipment manufacturershave coined acronyms for their own particular use

(Box 1).We now examine the standard SE sequence

that is at the heart of most conventional MRI.

Other pulse sequences, such as IR and uid-attenuated inversion recovery (FLAIR), are ex-plained. Gradient-recalled echo and limited ip

angle techniques that are variations of the SEpulse are also introduced. I limit discussion ofthe pulse sequences in this article based only ontheir ability to discriminate dierent tissue signal

characteristics, including T1 and T2 relaxation.The length of this primer does not allow one tocover the breadth of MRI pulse sequences.

Modications of these sequences are also used tomeasure ow, phase, diusion, and perfusion aswell as to reduce artifacts and perform functional

imaging. The interested reader is referred to thetextbook Neuroimaging, Clinical and PhysicalPrinciples [31].

Table 5 is a summary of some of the mostcommon pulse sequences used in MRI today. Abasic understanding of the pulse sequences used togenerate signals in the NMR experiment is neces-

sary, because this lies at the heart of the datarecorded. The pulse sequence can be thought of inthree phases:

1. A preparation pulse to excite the tissue. The

manner in which the tissue is excited, whetherit is a short ip angle or large ip angle, hasa signicant impact on T1 contrast.

2. A time interval between excitation of thetissue and acquisition of the data. This is theperiod during which dephasing (T2 relaxa-tion) occurs. A longer time increases T2 effect.

3. The overall time between data sampling, orTR. A long TR allows samples to recover,minimizing T1 contrast, whereas a short TR

accentuates T1 contrast.

Spin echo pulse

A SE sequence is established as follows. RFenergy is given to the system at the Larmour

frequency with enough intensity to ip the mag-netic vector into the x-y plane. This is the so-

20 GIBBYBox 1. Acronyms

3D FASTER Three-dimensional field echo acquisition with a short repetition time andecho reduction

3D GRE Three-dimensional gradient echo3D MPRACE Three-dimensional magnetization prepared rapid gradient echoADC Apparent diffusion coefficientBASE Basis imaging with selective inversion-preparedbEPI Blipped echoplanner imagingBMS Bulk magnetic susceptibilityBOLD Blood oxygenation level-dependent contrastBOSS Bimodal slice select radiofrequency pulseBP MR Biphasic MRIBW BandwidthCBF Cerebral blood flowCBV Cerebral blood volumeCE-FAST Contrast-enhanced Fourier acquired steady-state techniqueCNR Contrast-to-noise ratioCP Cross-polarizationCPMG Carr-Purcell-Meiboom-Gill (measurements of T2)CSF Cerebrospinal fluidCSMEMP Contiguous slice multiecho multiplanarDIGGEST Direct imaging of local gradients by group echo selection tomographyDISE Driven inversion spin echoDMSSFP Double-mode steady-state free precessionDOPING Double pulse interfaced echo imagingDPSF Diffusion perfusion snapshot flashDSC Dynamic susceptibility contrastDWI Diffusion-weighted imagingEPC Echo phase correctionEPI Echoplanar imagingEPISTAR Echoplanar imaging and signal targeting with alternating radiofrequencyETL Echo train lengthFAcE Free induction decay acquired echoesFAISE Fast acquisition interleaved spin echo (which is the same as fast spin echo)FAST Fourier-acquired steady-state techniqueFATS Fat-suppressed acquisition with echo times and real times shortenedFC Flow compensationFE Field echo, frequency encodeFEER Field even echo rephasingFFE Fast field echoFFF Fast Fourier flowFFP Fast Fourier projectionFID Free induction decayFIRFT Fast inversion recovery Fourier transformFISP Fast imaging with steady-state precessionFLAG Flow-adjusted gradientsFLAIR Fluid attenuation inversion recoveryFLASH Fast low-angle shotfMRI Functional MRIFONAR Field focusing nuclear magnetic resonanceFOV Field of view

21BASIC PRINCIPLES OF MAGNETIC IMAGINGFR Frequency encodeFSE Fast spin echo (turbo spin echo)FT Fourier transformFWHM Full-width at half-maximumG GaussGARP Globally optimized alternating phase Rectangular pulseGATORCIST Respiratory gated imagingGd GadoliniumGINSEST Generalized interferography using spin echoes and stimulated echoesGMN Gradient moment nullingGMR Gradient moment rephrasingGRASE Gradient spin echoGRASS Gradient acquisition in steady stateGRE Gradient echo imagingGREAT Ghost reduction by equalized acquisition tripletsGROPE Generalized compensation for resonance offset and pulse length errorsHASTE Half-Fourier acquisition single-shot turbo spin echoIR Inversion recoveryIR-EPI Inversion recovery echoplanar imagingIVIM Intra voxel incoherent motionLFA Limited flip angleMAST Motion artifact suppression techniqueMBEST Modulus blipped echoplanar single-pulse techniqueMBS-MRA Minimum basis set magnetic resonance angiographyMEMP Multiecho multiplanarMESS Multiple echo single shotmFISP Mirrored fast imaging with steady-state precessionMIP Maximum intensity projectionMOTSA Multiple overlapping thin slab acquisitionMPGR Multiplanar gradient recalledMPIR Multiplanar inversion recoveryMPRAGE Magnetization prepared rapid gradient echoMR Magnetic resonanceMRA Magnetic resonance angiographyMRI Magnetic resonance imagingMS-EPI Multishot echoplanar imagingMSIT Multiple slab imaging techniqueMT Magnetic transferMTC Magnetization transfer contrastMTR Magnetization transfer ratioMTSA Multiple thin slab acquisitionNEX Number of excitationsNMR Nuclear magnetic resonanceNSA Number of signal averagesPAIR Partial volume-sensitized inversion recoveryPC Phase contrastPE Phase encodingPEDD Proton-electron dipole dipolePEG Phase encode groupingPGSE Pulsed gradient spin echoPIETIR Prolonged inversion and echo time inversion recovery

(continued on next page)

22 GIBBYPOMP Phase-ordered multiplanarPPG Peripheral pulse gatingPPM Parts per millionsPRE Proton relaxation enhancementPRFT Partially relaxed Fourier transformPSIF Mirrored fast imaging with steady precessionPT2 Preferential T2QCSI Quantitative chemical shift imagingQMRI Quantitative MRIQUIPSS Quantitative imaging of perfusion using a single subtractionRACE Real time acquisition and evaluation of motionRAM FAST Rapid acquisition matrix Fourier acquired steady-state techniqueRARE Rapid acquisition relaxation enhancedRARE Rapid acquisition with refocused echoesRASE Rapid acquisition spin echoRBC Red blood cellrCBF Regional cerebral blood flowRF RadiofrequencyRF-FAST Radiofrequency Fourier-acquired steady-state techniqueROI Region of interestROPE Respiratory ordered phase encodingRUFIS Rotating ultrafast imaging sequenceSAAV Simultaneous acquisition of artery and veinSAR Specific absorption rateSAT Saturation pulseSD Standard deviationSE Spin echosEPI Spiral echoplanar imagingSIMUSIM Simultaneous multislice imagingSIP Saturation inversion projectionSMART Simultaneous multislice acquisition using rosette trajectoriesSmaRT Simulataneous multislice acquisition with arterial-flow taggingSMI Simulataneous multislice imagingSNR Signal-to-noise ratioSPACE Spatial and Chemical-shift encoded excitationSPAMM Spatial modulation of magnetizationSPECT Single photon emission computed tomographySPGR Spoiled gradient recalled (spoiled gradient acquisition in steady state)SPIR Selective population inversion recoverySS Slice select gradientSSFP Steady-state free precessionSSP Section-sensitivity profileSTE Stimulated echoSTIR Short tau (inversion time) inversion recoverySTREAM Suppressed tissue with refreshment angiography methodT TeslaT2 FFE T2 fast field echoT2 PEDD T2 proton electron dipole dipole interactionT2 PRE T2 proton relaxation enhancementTCF Time correlation functionTD Trigger delay

TE Time delay between excitation and echo maximum

the x-y plane in a coherent fashion. Over a shortp

vdo

dFim

cinon

hthm

thish

cOa

ath9e

[3o

The entire pulse sequence is repeated manycal experiment. Several averages or

tations (NEXs) may be obtained tonal-to-noise ratio. Multiple phase-are taken to achieve spatial local-

me for which the pulse sequence ised the TR, or time of repetition.the entire pulse sequence. Image

unction of the timing parametersR, and tissue-specic propertiesage contrast has some eects fromroton density; therefore, sequences

as being weighted toward a given

weighting

st the consequences of altering the

at the TE is taken to be as short asduce T2 or dephasing eects. Ifsed, all the magnetization will have

e z-axis. At the start of the nextwill be available for deection intohis gives maximum signal, and the

intensities of tissues are based notT1 relaxation characteristics but onrogen there is (ie, proton density),

se

xcitdsrem

ntra

tic I

uennsfo

23BASIC PRINCIPLES OF MAGNETIC IMAGINGeriod, the individual nuclei comprising the net

ector drift out of phase. The signal rapidlyecays as an FID. If the receiver coil were turnedn at this time, a sinusoidal wave of rapidly

ecreasing intensity would be produced (ie, anID). A certain time later (1100 milliseconds inaging), a second RF pulse is given, which now

orresponds to a 180( pulse. The vectors areverted, which causes them to spin in thepposite direction. Fast-spinning protons areow behind the slower protons, and phase co-

erence can be re-established for those protonsat became out of phase because of staticagnetic eld inhomogeneities. Fig. 25 displays

e SE pulse sequence in which the magnetizationdeected into the x-y plane, loses phase co-

erence and is then inverted by a 180( pulse,hanges rotational direction, and is rephased.ver a period equal to the time between the 90(nd 180( pulses, phase coherence is re-establishednd signal is generated as an echo. The data areen acquired. The total elapsed time from the0( pulse to the echo is called the TE, or time tocho. This is the prototype SE, or Hahn echo

2], described only a few years after the discoveryf NMR.

times in a typi

number of exciincrease the sigencoding steps

ization. The tirepeated is callFig. 26 shows

contrast is a fchosen, TE, T[3335]. All imT1, T2, and p

are designatedparameter.

Proton density

Consider r

TR. Assume thpossible to rea long TR is u

returned to thpulse train, itthe x-y plane. T

relative signalon the relativehow much hydTEI TE interleavedTFE Turbo field echoTI Time following inversion pulTMR Topical magnetic resonanceTOF Time of flightTONE Tilt optimized nonselective eTOSS Total suppression of sidebanTPPI Time-proportional phase incTR Time to repetitionTRICKS Time-resolved imaging of coTSE Turbo spin echoTSR Total saturation recoveryTurbo FLASH Turbo fast low-angle-shotURGE Ultra rapid gradient echoUSPIO Ultra small superparamagneVAS Variable angle spinningVEMP Variable echo multiplanarVENC Velocity encoding valueVIGRE Gradient echoVINNIE Velocity encode cine imagingVOI Volume of interestVPS Views per segmentWATERGATE Water suppression pulse seqWEFT Water-eliminated Fourier traation

entation

st kinetics

ron oxide

cerm

Table 5

Common pulse s

Acronym P Range Flip TI Contrast eect

SE S 200600 90( None T1W20004000 90( Proton20004000 90( T2W

CPMG M 20004000 90( None Proton and T2W

GRASS G 1050a 45( None Proton/steady state1050 2000 milliseconds

radient echo Short 220 Short

Long 540 Short

Short 510 Long

poiled grass Short 210 Short

version

recovery

Minimum 1020 Long

Short >2000 milliseconds

hort time

inversion

recovery

Long 50120 Long

>3000 milliseconds

luid-attenuated

inversion

recovery

Long 80200 Very long

150 milliseconds 6000 milliseconds

n ip angle.

25BASIC PRINCIPLES OF MAGNETIC IMAGINGas illustrated in Fig. 27. Proton density contrast isoften misunderstood or ignored. Wehrli et al [36]demonstrated that most of the contrast seenbetween gray and white matter on T2W SE

sequences can be ascribed to dierences in protondensity: gray matter has more water protons thanwhite matter. Furthermore, to achieve maximum

tissue contrast, the selection of pulse sequence tobe used is highly dependent on hydrogen spindensity. As the ratio of spin densities increases

between two substances, SE becomes a betterpulse sequence than IR [37].

T1 weighting

Suppose, however, that only a short timeis allowed for the magnetization to recover tothe z-axis. Only those substances that have

extremely short T1s will have achieved their fullpotential magnetization before being pulsed with

sequence then becomes T1W, allowing for dier-ential intensities to be observed between substan-ces that have diering T1 values. The equationdescribing just the recovery of longitudinal mag-

netization as a function of time is: Magnetiza-tion =Mz (1 eTR/T1), where Mz is the totalnet magnetic vector in the z-axis before 90(excitation, TR is repetition time, and T1 isa constant for each tissue. The time for completerelaxation is innity. For 99% recovery, one must

wait 4.6 times T1, as shown previously in Fig. 15.Obviously, most of the relaxation occurs withinthe rst 2.0 times T1. Dierent tissues and sub-

stances have characteristic relaxation rates specicto that individual material. Furthermore, as wasdiscussed earlier, T1 relaxation is also dependenton magnetic eld strength. To optimize tissue

contrast between voxels containing elements ofdierent T1 relaxation values, one should thusknow what the relaxation rate for a given tissue is.

Fig. 25. Magnetic vector diagram of standard spin echo sequence. The incoherent precessing vectors are brought into

coherence, and the net vector is tipped into the x-y plane. A free induction decay (FID) occurs. After a short time, the

vectors begin to dephase. A 180( radiofrequency (RF) pulse is then applied, inverting the vectors and reversing theirdirection. After a period of time has elapsed (echo time [TE]), the vectors rephase and an echo is produced.a repeat pulse, as illustrated in Fig. 28. The pulse In Fig. 29, a family of curves of tissues with

26 GIBBYFig. 27. Proton-weighted sequence. A 90( radiofrequency pulse is given. A long time (repetition time [TR]) elapses, andall the tissues (A, B, and C) relax to the ground state. When the next 90( pulse is given, 100% of the magnetization isavailable to tip again into the x-y plane. Therefore, maximum signal is achieved. Only if the materials have a dierent

proton density (ie, quantity of available mobile hydrogen) is there a dierence in signal between the three tissues.dierent T1 relaxation values is illustrated todiscriminate between fat and white matter whichhave short T1s, a short is used. To achieve

maximal contrast between tumor and CSF, thena longer TR would be best. In general, one shouldselect a TR close to the T1 value of the tissue of

interest. This ensures the widest possible separa-tion between tissues with close T1 values.

T2 weighting

The SE sequence can also be used to acquireT2W data. In this instance, the TR between pulses

is set quite long so that as much of the longitudinalmagnetization as possible can recover (ie, no T1eects). The time before data acquisition is nowlengthened, however. The 180( pulse is given atamuch later time, allowing for increased dephasingto occur. Only those tissues with long T2s (ie, thosethat dephase very slowly) have enough residual

phase coherence available so that when the 180(pulse is applied, they can be brought back intophase. Because of phase losses incurred from non-static magnetic eld inhomogeneities, only a frac-

tion of the initial magnetization vector can berecovered. By necessity, less and less signal is ac-quired as TEs are lengthened and images become

noisier. Fig. 30 illustrates this pulse sequence.Fig. 26. Standard spin echo (SE) timing diagram. This gure demonstrates the radiofrequency (RF) pulse timing and

associated signal from a standard SE sequence. A 90( pulse is given, followed by a 180( RF refocusing pulse at 12 echotime (TE). A period of time (repetition time [TR]) then elapses, and the entire process is repeated. An FID (free induction

decay) occurs after the 90( pulse but the signal is actually acquired at the echo.

27BASIC PRINCIPLES OF MAGNETIC IMAGINGFig. 29. Relaxation curves for dierent tissues, A through E. The optimal time to discriminate between fat and white

matter would be at 318 milliseconds. In other words, a short repetition time (TR) is best to discriminate between tissues

of short T1 values. A longer TR would be better to discriminate between tissues of longer T1 values, such as tumor and

cerebrospinal uid; in this case, 1883 milliseconds at 1.5-T eld strength. CSF, cerebrospinal uid.The rst part of this pulse sequence is exactly likeSubstances that have prolonged T2 values include

free water, such as CSF, edema, cysts, and mostpathologic processes in which tissue injury hasoccurred. T2W images, although having a lower

signal-to-noise ratio than T1W images, are stillthe most useful for diagnostic neuroimaging [36].

Carr-Purcell-Meiboom-Gill sequence

The CPMG sequence [38,39] is a commonlyused variation of the SE pulse sequence. In fact,most T2W SE sequences use this technique to

acquire proton and T2W images simultaneously.

Fig. 28. T1-weighted pulse sequence. A 90( radiofrequency pulse is given, and a short time elapses before repeating theprocess. Tissue A with a short T1 has relaxed to the ground state, giving a maximum vector when reipped into the x-y

plane. Tissue C with a long T1 has not relaxed to the ground state, however. When the tissue is given a new 90( pulse,only a small vector is produced, creating substantially less signal intensity. In this pulse sequence, tissues with a short T1

relaxation time (TR) are the brightest.

28 GIBBYthe SE sequence that we have previously de-

scribed. Soon after the rst echo is produced,there is rapid dephasing of the proton spins andsignal is lost. A second or third 180( RF pulse canthen be applied, which reverses the spinningvectors and brings them back into phase. Becauseof T2 relaxation caused signal loss in the tissue, all

of the signal cannot be rephased. Thus, our signalprogressively gets smaller and smaller with eachecho. This echo train, as illustrated in Fig. 31, isa curve tted by plotting the maximum signal

intensities at each echo and represents the true T2relaxation curve for the tissue. The rapid T2 decayfor the FID of each echo is the result of true tissue

T2 and dephasing from static magnetic eldinhomogeneity. Together, these are called T2*.

Inversion recovery

There are several types of inversion recovery(IR) pulses, which are variations on a theme buthave signicantly dierent appearances in terms

of image contrast [40] and can be used for a widevariety of clinical applications. These are dis-cussed separately as conventional IR, short-time

inversion recovery (STIR), and uid-attenuatedinversion recovery (FLAIR).

Conventional inversion recovery

In the usual IR pulse sequence [34], a 180(pulse is given, which rotates the magnetizationvector into the negative z-direction, as shown inFig. 32. Note that this requires twice the RF power

Fig. 30. T2-weighted spin echo pulse. A long repetition time is used such that all the magnetization is available before

tipping into the x-y plane. The time before the 180( refocusing pulse is relatively long (ie, a long echo time [TE] is used).For tissues with long T2 relaxation, most of the signal remains coherent. For those with a short T2 relaxation time (TR),

only a small amount of the signal is recovered with the 180( radiofrequency (RF) pulse.

29BASIC PRINCIPLES OF MAGNETIC IMAGINGand that it also requires a longer time to recoverto the steady-state positive z-direction. This isgoverned by the exponential decay time constant,T1. If TR is assumed to be extremely long relative

to T1, the equation describing T1 relaxation can beshown to be: Signal = So (1 2eTI/T1), where So isthe total z-component of the magnetic vector, TI is

the time to inversion, T1 is the familiar relaxationconstant (which is tissue dependent) [33], and TRis repetition time.

Fig. 33 shows the typical way in which thesignal intensity from an IR pulse sequence isplotted as a function of time. After inversion, asthe magnetization vectors begin returning to the z-

axis, those with short T1s do so rst. At a timecalled TI, a 90( pulse is given. If this is performed

Fig. 31. Carr-Purcell-Meiboom-Gill sequence. This sequence applies a series of refocusing 180( radiofrequency (RF)pulses with repeated echoes. A curve of signal decay can be traced, giving