introductiontomagnetic resonance imaging for basel ... · standing of mri by introducing the basics...

Introduction toMagneticResonance Imaging forNeurologists

Ernst-Wilhelm Radue, MD; Matthias Weigel, PhD; Roland Wiest, MD;Horst Urbach, MD

ABSTRACTPurpose of Review: In neuroradiology, highly sophisticated methods such as MRIare implemented to investigate different entities of the central nervous system andto acquire miscellaneous images where tissues display varying degrees ofcharacteristic signal intensity or brightness. Compared to x-ray, CT, and ultrasound,MRI produces clearer images of tissues, body fluids, and fat. The basics of MRI maybe unknown to neurologists; this article introduces MRI physics, techniques, andinterpretation guidelines.Recent Findings: This article discusses the basics of MRI to provide clinicians withthe scientific underpinning of MRI technology and to help them better understandimage features and improve their diagnosis and differential diagnosis by combiningMRI characteristics with their knowledge of pathology and neurology.Summary: This article will help neurologists deepen their knowledge and under-standing of MRI by introducing the basics of MRI physics, technology, imageacquisition, protocols, and image interpretation.

Continuum (Minneap Minn) 2016;22(5):1379–1398.

INTRODUCTIONThis article provides a short, butcomprehensive, overview of MRI forneurologists, including the physicsunderlying this imaging modality, thebasics of image acquisition, and theapproach to imaging interpretationwith MRI.

A PRIMER ON MRI PHYSICSMRI is a noninvasive imaging modalitythat provides excellent soft tissuecontrast for normal and pathologicstructures of tissues such as cartilage,muscles, brain, and spinal cord as wellas fat and body fluids. Medical imagingmodalities often produce image con-trasts that are related to the differentabsorption and reflection of waves inbody tissues (eg, x-ray, CT, ultra-

sound). However, MRI produces im-ages that are related to the magneticproperties of, and molecular interac-tions within, the tissues under obser-vation. Since these interactions arecomplex and manifold, MRI has be-come a powerful and versatile imagingmodality that can distinguish a largenumber of very different tissue contrasts.

Important Physical FeaturesThat Create MRI ImagesThe physics of MRI can be understoodbest by following a number of stepsand thereby investigating the underly-ing physics and technology. Thesesteps also reflect the physics of anMRI measurement cycle.

Polarization. The human body con-sists of approximately 1027 molecules,

Address correspondence toDr Ernst-Wilhelm Radue,Biomedical Research andTraining, University Eye HospitalBasel, Mittlere Strasse 91,CH-4031 Basel, Switzerland,[email protected].

Relationship Disclosure:

Dr Radue has served as aconsultant for Neurologischeund PsychiatrischeUniversit.ts Klinik Basel,has received personalcompensation for speakingengagements from NovartisAG and Sanofi Genzyme, andreceives royalties fromSpringer-Verlag GmbH.Dr Weigel has receivedpersonal compensation forspeaking engagements fromthe European Society forMagnetic Resonance inMedicine and Biology, theMagnetic Resonance CompactUniversity Hospital Erlangen,and VereinigungSudwestdeutscherRadiologen undNuklearmediziner androyalties from the UniversityMedical Centerof the University of Freiburgfor a patent for dynamicmotion correction in MRI.Dr Wiest receives research/grant support from the SwissNational Foundation. DrUrbach serves as coeditor ofClinical Neuroradiologyand on the editorial board ofNeuroradiology and hasreceived personal compensationfor speaking engagements fromBracco, Stryker, and UCB SA.

Unlabeled Use of

Products/InvestigationalUse Disclosure:Drs Radue, Weigel, Wiest, andUrbach report no disclosures.

* 2016 American Academyof Neurology.

1379Continuum (Minneap Minn) 2016;22(5):1379–1398 www.ContinuumJournal.com

Review Article

Copyright © American Academy of Neurology. Unauthorized reproduction of this article is prohibited.

mailto:[email protected]

which is an unfathomably large num-ber. Each molecule consists of a fewto thousands of atoms. Many nuclei ofthese molecule-bound atoms rotatearound their axis; they have a so-calledspin.1Y3 Spinning nuclei always gener-ate a nuclear magnetic (dipole) mo-ment. Thus, each spinning nucleusbehaves like a tiny bar magnet with amagnetic north pole and south pole. Ithas become common in MRI literatureto refer to these nuclear magnets in asimplified way as spins.1Y3 Because oftheir high natural abundance in thehuman body and because they have thestrongest magnetic moment, the nucleiof hydrogen atoms are commonly usedfor clinical MRI. A hydrogen nucleusconsists of only one proton. There-fore, the following discussion focuseson spins of hydrogen nuclei/protons.

Although the human body containsroughly 1027 magnetic hydrogen nu-clei, ie, spins, humans do not seem tobe magnetic. Why? Because all thespins are randomly orientated in thebody; they cancel each other, and there-fore their net magnetization is 0.1Y3

The basic idea of MRI is to put thepatient into a very strong and staticmagnetic field, the so-called main fieldB0. This static B0 field is generated bythe MRI scanner and is approximately30,000 to 140,000 times stronger thanthe earth’s magnetic field. As a conse-quence, the spins begin to orient inspace and to align with the B0 field.However, compared to compassneedles in the earth’s magnetic field,spins align not only parallel but alsoantiparallel to the external static B0

field.1Y3 The numbers are almostequal; only a tiny excess of spins isadditionally aligned parallel to the B0

field, leading to a tiny but measurablemacroscopic net magnetization, orM0. Since this effect of so-calledpolarization is so weak, MRI is animaging modality that permanently

suffers from low signal (and low signalto noise ratio, as described later inthis article).

Excitation and precession. Initially,at rest, the net magnetization residesparallel to the B0 field in the equi-librium position; this is called theequilibrium magnetization M0.

1Y3 How-ever, an MRI scanner is able to gener-ate a second magnetic field, B1, thatoscillates in the radiofrequency range.The spins are able to absorb energyfrom this B1 field, and, thus, they aretipped or flipped out of the equilib-rium position. This process is calledexcitation; the spins and, accordingly,the netmagnetization are excited.1Y3 Theangle by which the spins/magnetizationare rotated is termed the flip angle. Flipangles are usually between 0 degreesand 90 degrees.

As soon as the spins leave theequilibrium position, they start toundergo a rotating motion aroundthe external B0 field called precession,much like a spinning top under theeffect of gravity. The precession fre-quency of the spins is depicted by theLarmor equation.1Y3 The Larmor equa-tion states that the precession fre-quency is directly proportional to thestrength of the magnetic field. Thus,the higher the magnetic field, thehigher the precession frequency (ie,the faster the spins rotate). Contraryto this, the lower the magnetic field,the lower the precession frequencyand the slower the spins rotate.

The basic statement of the Larmorequation would not be of such greatimportance in MRI if all spins experi-enced the same magnetic field asdetermined by the nominal value ofthe scanner’s B0 field (eg, 1.5T or 3T)and therefore automatically had thesame Larmor frequency. This is notthe case for several reasons. First, theB0 main field of an MRI scanner cannever be built perfectly constant in

KEY POINT

h MRI produces imagesthat are related to themagnetic properties of,andmolecular interactionswithin, the tissuesunder observation.

1380 www.ContinuumJournal.com October 2016

Introduction to MRI


space as it has spatial inhomogenei-ties. Second, human body tissues alsogenerate a tissue-characteristic magne-tization called susceptibility, whichdistorts the external B0 field. Hence,the spin’s Larmor frequency slightlychanges with position due to B0

inhomogeneity and tissue susceptibil-ity effects.1Y3 The third reason is thatan MRI scanner can generate spatiallyvarying magnetic fields that are usedfor spatial encoding. These so-calledgradient fields are discussed later inthe article.

So far, it has been roughly said thatthe B1 excitation field oscillates in theradiofrequency range; however, to bemore accurate, the B1 oscillationfrequency has to match quite preciselythe Larmor precession frequency of thespins. The two frequencies have to bein resonance. Otherwise, the spinscannot absorb the radiofrequency en-ergy, and they are not excited.1Y3 Agood analogy for this observation ofmandatory resonance is a child’s swing.The child must move synchronously orin resonance with the swing to increaseits amplitude. The modality’s full name,written out as nuclear magnetic reso-nance imaging, directly originatesfrom this effect.1Y3

Magnetization is a vector quantity,ie, it possesses a magnitude and adirection (Figure 1-1). As has alreadybeen pointed out, the initial directionof the net magnetization is parallel tothe external B0 field. By definition, thisis the z-axis of the scanner’s coordi-nate system; it is also called thelongitudinal direction. Orthogonal tothe longitudinal z-direction is thetransverse plane, spanned by theorthogonal x-axis and y-axis. In MRI,any magnetization vector is de-composed into a longitudinal and atransverse component (Figure 1-1).

Measurement with radiofrequencycoils. After excitation, transverse mag-

netization components occur and pre-cess in the x-y-plane. Generally,moving or rotating magnetic momentsor vectors induces electrical currentsin electrical coils. Thus, the precessingexcited spins induce an electricalvoltage in radiofrequency receivercoils placed around the human body,similar to a generator producingelectricity.1Y3 Since the effect is veryweak, the measured radiofrequencysignal is tiny and must be amplified bya dedicated radiofrequency receiver/amplifier chain. However, the radio-frequency coils and radiofrequencyamplifier chain also measure andgenerate noise, which is usually ofsimilar magnitude to the desired ra-diofrequency signal. For this reason,MRI inherently suffers from a low signalto noise ratio. If the MRI imagesbecome too noisy (ie, if the signal tonoise ratio gets too low), diagnosis isno longer feasible.

Modern radiofrequency coils forMRI are usually close to the patient’sbody to improve the reception of the

FIGURE 1-1 In MRI, any magnetization vector M(orange arrow) is decomposed into alongitudinal component Mz parallel to

the z-axis (blue arrow) and into a transversecomponent Mxy (green arrow) in the x-y-plane.



radiofrequency signal. Since the mea-sured noise is directly correlated tothe sensitivity volume of the radio-frequency receiver coil, small radio-frequency coils produce a highersignal to noise ratio than the scanner’slarge body coil. Furthermore, modernphased-array coils take advantage ofthis; they consist of a matrix of severalsmall radiofrequency coils of highsignal to noise ratio, combining themto be sensitive to a larger body volume.At the end of the radiofrequency re-ceiver chain, the signal is furtherpostprocessed by a computer to re-construct the magnetic resonance(MR) image.1Y3

Relaxation. Nature always strivesfor equilibrium, ie, for a uniformdistribution of energy. Consequently,in MRI, the excited spins want to getrid of their energy such that the netmagnetization resides as a pure longi-tudinal magnetization in the equili-brium position again. This effect iscalled relaxation.1Y4 Two differenttypes of relaxation occur in MRI, whichis another reason magnetization vec-tors are always regarded as a longitu-dinal magnetization component and atransverse magnetization component.

With the equilibrium magnetizationbeing excited, the longitudinal com-ponent is reduced (flip angle between0 and 90 degrees) or becomes exactly0 (a 90-degree flip angle). Then theprocess of longitudinal relaxationimmediately starts: the longitudinalmagnetization grows back to its nor-mal size in the equilibrium state.However, this is not an immediateprocess, because the spins need timeto transfer their excess of energy tothe environment in the tissue. To bemore accurate, the speed of thelongitudinal relaxation is specific for agiven tissue and takes place with acharacteristic relaxation time T1.1Y4

Furthermore, this T1 relaxation corre-

sponds to an exponential asymptoticcurve (Figure 1-2). Thus, T1 does notdirectly depict the time when the re-laxation is completed, but is a mea-sure of the speed of the relaxation orrecovery of the longitudinal magneti-zation component (Figure 1-2).

T1 relaxation times are not onlyhighly tissue characteristic, but theyalso depend on the B0 field strengthof the scanner.1,2,4 They also alter withstructural changes of the tissue, whichmakes T1 highly sensitive for patho-logic changes of body tissue; T1-weighted MRI uses T1 relaxation timesto generate fundamental images withcharacteristic tissue signal intensitiesor tissue brightness (Figure 1-2). T1relaxation times are roughly betweenhalf a second and a few seconds forliquids. As a rule of thumb, the moreaqueous a tissue is, the higher the T1is. Fat has short T1.

A second relaxation takes place thatleads to a decay or relaxation of thegenerated transverse componenttoward 0, which can be considerablyfaster than the recovery of the longi-tudinal component.1Y4 Remember thatthe macroscopic net magnetization isthe sum of a multitude of spins. Foreach spin, the Larmor equation holdstrue. Since the molecules and spinsare in permanent thermal motion ona microscopic scale, their magneticmoments interact with each other.To put it simply, a given spin influ-ences the local magnetic field theneighboring spins “see”; accordingly,their Larmor precession frequencieschange constantly. As a consequence,the precessing transverse magnetiza-tion components of the spins get outof sync or out of phase.1Y4 The in-dividual components fan out in thetransverse plane, leading to a decay ofthe transverse net magnetization withthe characteristic and tissue-specificrelaxation time T2 (Figure 1-3). The T2

KEY POINT

h T1-weighted MRI usesT1 relaxation times togenerate fundamentalimages with characteristictissue signalintensities or tissuebrightness. As a rule ofthumb, the moreaqueous a tissue is, thehigher the T1 is. Fat hasshort T1.


Introduction to MRI


FIGURE 1-3 Transverse relaxation or T2 decay curves of the (transverse) magnetizationcomponent Mxy shown for two tissues with different T2 but the same protondensity. In this example, tissue 1 has a short T2 (eg, body fat, orange curve),

whereas tissue 2 has a long T2 (eg, CSF, blue curve). Based on such different decay curves,T2-weighted images can be acquired. Three representative examples are shown, withincreasing echo time from left to right, leading to an increase in T2 weighting from left to rightaccordingly. As is typical for a T2-weighted brain image, gray matter is brighter than whitematter because of the longer T2. CSF has the longest T2 and is therefore very bright.

FIGURE 1-2 Longitudinal relaxation or T1 recovery curves of the (longitudinal) magnetizationcomponent Mz shown for two tissues with different T1 but the same protondensity. In this example, tissue 1 has a short T1 (eg, body fat, orange curve),

whereas tissue 2 has a long T2 (eg, CSF, blue curve). Based on such different recovery curves,T1-weighted images can be acquired; two representative examples are shown. The left imagewas acquired with a shorter repetition time than the right image, leading to a stronger T1weighting. As is typical for a T1-weighted brain image, white matter is brighter than graymatter because of the shorter T1. CSF has the longest T1 and is therefore very dark.



relaxation is an exponential decay;thus, similar to T1, T2 does not di-rectly depict the time when therelaxation is completed but is a mea-sure of the speed of relaxation ordecay (Figure 1-3).

T2 relaxation times are approxi-mately between 50 milliseconds and100 milliseconds for aqueous softtissues and roughly 1 to 2 secondsfor liquids. T2 also alters with struc-tural or metabolic changes of thetissue, which makes it highly sensitivefor pathologic changes of body tissue;T2-weighted MRI uses T2 relaxationtimes to generate fundamental imageswith characteristic tissue signal inten-sities or tissue brightness (Figure 1-3).

Because of the random nature ofthe thermal motion, the spin interac-tions are random, and, thus, the T2decay is irreversible.1Y4 However, anadditional, reversible effect leads tofurther decay of transverse magnetiza-tion. Inhomogeneities of the scanner’sB0 field and varying susceptibility ef-fects of the tissues, as discussed earlierin the article, also change the Larmorfrequency according to the Larmorequation. These effects are fixed inspace on a macroscopic scale and are,therefore, reversible by using so-calledspin echoes.4 The combined relaxationof the irreversible T2 part and thereversible part is termed T2*.1Y5

Spatial EncodingSo far, basic MR physics, signal gener-ation, and measurement as well asrelaxation effects have been discussed.However, to reconstruct a full image,the measured MR signal needs to beassigned to the corresponding loca-tion in the patient’s body where it wasgenerated; spatia l encoding isneeded.1Y5 To achieve this, so-calledgradient coils are employed that gen-erate a linear magnetic field in space.In accordance with the Larmor equa-

tion, the Larmor frequency alsochanges linearly along the gradient’sdirection accordingly. This effect isexploited in three ways for spatialencoding. First, during radiofrequencyexcitation, a gradient can be turnedon, the so-called slice selection (two-dimensional acquisition) or slab selec-tion (three-dimensional acquisition)gradient.1Y5 Consequently, accordingto the resonance condition mentionedearlier in the article, only spins areexcited (and therefore generate signal)where the linearly changing precessionfrequencies in space match the rangeof radiofrequency excitation frequen-cies: This defines the desired slice orslab. Second, during the measurement,a gradient can be turned on, the so-called read or frequency-encodinggradient.1Y5 Then, the spin locationsalong this read gradient are encodedinto the signal via the spins’ spatiallydependent Larmor frequency. Third,between excitation and measurement,a gradient can be turned on, the so-called phase-encoding gradient.1Y5

Thus for a duration of time, the spinsalong this phase-encoding gradientprecess with different Larmor frequen-cies and therefore acquire differentphases. As phases, the different anglesof transverse magnetization compo-nents in the x-y-plane are indicated(Figure 1-1). Repetitive measurementcycles with different phase-encodinggradients allow encoding spin locationsinto phase changes. Generally, threegradient coils also reside within thescanner, one each for the x-, y-, andz-direction.

The measured signal containingthe spatial encoding is stored into araw data space called k-space. After-ward, the main step in image recon-struction is to make use of the Fouriertransform. It facilitates decoding ofthe k-space data such that it as-signs signal intensities to their correct

KEY POINT

h T2 alters with structuralor metabolic changesof the tissue, whichmakes it highly sensitivefor pathologic changesof body tissue;T2-weightedMRI uses T2relaxation times togenerate fundamentalimages withcharacteristic tissuesignal intensities ortissue brightness.


Introduction to MRI


location again, resulting in the re-constructed image.

The complexity of the spatialencoding and the MR signal genera-tion makes MRI prone to errors inthese procedures, eg, caused by B0

inhomogeneities, susceptibility ef-fects, body motion, or wrong parame-ter settings at the scanner. Sucheffects can manifest in the image asartifacts, ie, structures are displayedthat do not reflect the reality, or theimage may be blurred. Identifying andavoiding such artifacts is of greatimportance in MRI.

MRI SequencesGenerally, radiofrequency excitation,the switching of spatial encoding gra-dients, waiting times for relaxation,and signal measurements follow anexact and predefined timing patternthat is called the (MRI) sequence.Some parts of the sequence timingmay be influenced by the user. Theexact sequence timing pattern deter-mines the characteristic brightness ofthe different tissues in the reconstructedimages due to proton density and T1and T2 weighting.

Dozens, hundreds, or perhapsthousands of different MRI sequencesexist, depending on how the differenttypes and variants that were publishedwithin the past 40 years are countedand classified. In practice, MRI sequencesare frequently classified as being eithera gradient recalled echo (GRE) or a spinecho sequence.4,5 Basically, GRE se-quences use a radiofrequency excita-tion pulse, apply the spatial encoding,and directly measure the excited mag-netization, as described earlier in thearticle. Spin echo sequences apply asecond radiofrequency pulse after exci-tation and before themeasurement, theso-called refocusing pulse. It reversesor refocuses the B0 inhomogeneity andsusceptibility effects. Whereas spin

echo sequences are more robust insignal behavior, they are slower thanGRE sequences. A faster spin echovariant, the turbo or fast spin echosequence, generates more than oneecho per excitation and is one of theworkhorses in clinical MRI.4,5

Important MRI sequence parame-ters that can usually be influenced bythe user at the scanner are the echotime (TE) and the repetition time(TR). TE defines the time from theexcitation until the signal is measuredin the transverse plane.1Y5 This is ofimportance because the longer the TEdefined by the user, the longer thetissue-specific T2 decay takes place.Hence, the acquired image displays astronger T2 weighting. TR defines thetime between repetitive measurementcycles or excitations.1Y5 If the definedTR is short compared to the charac-teristic T1 relaxation times of the in-vestigated tissues, the correspondinglongitudinal magnetizations cannotfully recover before the next excita-tion, which represents a T1 weightingof the magnetization. As a result, theimage that is reconstructed is alsoT1 weighted.

T1 and T2 weighting are dominantor strong signal weightings in MRI andtherefore facilitate images with excel-lent tissue differentiation. Morpho-logic acquisitions with both low T1weighting and low T2 weighting usuallybring out the third fundamental MRIweighting: proton density weighting.1Y5

This weighting is proportional to thenumber of spins/protons in the tissue.Proton density weighting is weak com-pared to T1 and T2 weighting, partic-ularly in the brain.

Magnetization PreparationThe flexibility and versatility of tissuebrightness and appearance in MRI canbe increased manifold by employingadditional radiofrequency and gradient

KEY POINTS

h Errors caused by B0inhomogeneities,susceptibility effects,body motion, or wrongparameter settings onthe scanner can manifestin the image as artifacts(structures displayed thatdo not reflect reality), orthe image may beblurred. Identifying andavoiding such artifacts isof great importancein MRI.

h Important MRI sequenceparameters that canusually be influenced bythe user at the scannerare the echo time andthe repetition time.

h Echo time defines thetime from the excitationuntil the signal ismeasured in thetransverse plane. This isof importance becausethe longer the echo timedefined by the user, thelonger the tissue-specificT2 decay takes place.Hence, the acquiredimage displays a strongerT2 weighting.

h Repetition time definesthe time betweenrepetitive measurementcycles or excitations.

h If the defined repetitiontime is short compared tothe characteristic T1relaxation times of theinvestigated tissues, thecorresponding longitudinalmagnetizations cannotfully recover before thenext excitation, whichrepresents a T1weightingof the magnetization.As a result, the imagethat is reconstructedis also T1 weighted.



pulses. These magnetization prepara-tions introduce dedicated weightingsinto the initial equilibrium magnetiza-tion before the actual excitation pulseor make the already excited magneti-zation sensitive to some physicaleffects. Both have a direct impact onthe tissue brightness and appearancein the later reconstructed image.

Among important variants are in-version recovery, diffusion, f low veloc-ity, and perfusion preparations. Aninversion recovery preparation usesan initial 180-degree radiofrequencypulse to invert the longitudinal mag-netization along the z-axis. Afterward,characteristic T1 recovery takes placetoward equilibrium for all tissues;however, all longitudinal magnetiza-tions have one moment in time whentheir magnitude is 0, depending onthe tissue’s T1. If the excitationfollows in such a moment, the corre-sponding tissue appears dark in theimage (tissue nulling). Importanttypes are short tau inversion recovery(STIR), in which fat is dark, and fluid-attenuated inversion recovery (FLAIR),in which fluids such as CSF are dark.

By using particular gradient patterns,MRI sequences can also be made sen-sitive to motion effects, such as diffu-sion and flow in the tissue.6 Theseparticular gradient patterns alter thephase of transverse magnetizationcomponents in dependence of mo-tion. Diffusion represents water mo-tion in the tissue on a microscopicscale; it is determined by the tissuemicrostructure and also reflects metab-olism.6 The resulting apparent diffu-sion coefficient (ADC) is tissue-specificand is a measure of the mobility of thewater molecules in the tissue. Diffu-sional motion is exceptionally sensi-tive to pathologic tissue changes, and,thus, diffusion-weighted imaging (DWI)is extremely well suited as an earlymarker for stroke imaging. Since diffu-

sional motion is of a random nature,the introduced phase change is dif-ferent for each individual spin; hence,the net transverse magnetization isattenuated.6 By contrast, f low in avessel is a coherent motion; thus, theintroduced phase change for each in-dividual spin is the same and pro-portional to its speed. Consequently,f low velocities can be measured fromthe phase change of the net trans-verse magnetization.

Direct perfusion-weighted imagingis also feasible by administering con-trast agents to the patient. MRI con-trast agents cause a strong reductionof the local T1 and T2 relaxation timesin the blood and where the contrastagent molecules are able to diffuse,which may also depend on the whetherthe blood-brain barrier is intact. Addi-tional postprocessing techniques andmodels allow quantitative perfusionvalues to be derived. Prominent exam-ples of perfusion-weighted imaging aredynamic contrast-enhanced imaging(T1-weighted) and dynamic susceptibil-ity contrast (T2-weighted) imaging.Generally, apart from perfusion-weighted imaging, contrast agents mayadd further information about tissuedamage and are able to brighten upvessels in the body.

IMAGE ACQUISITION ANDMAGNETIC RESONANCEPROTOCOLSThe principal advantage of MRI is theability to investigate tissue-dependentresponses to a magnetic field in multi-ple planes without the use of ionizingradiation. Basic knowledge about thebrain’s anatomy and typical landmarksis mandatory for reading MRI.

To acquire the appropriate slicinglocalizer scans using fast echo, acquisi-tions in three orthogonal planes arerecommended. Axial imaging ac-quisitions should be placed in the

KEY POINTS

h T1 and T2 weighting aredominant or strongsignal weightings in MRIand therefore facilitateimages with excellenttissue differentiation.

h Diffusion representswater motion in thetissue on a microscopicscale; it is determined bythe tissue microstructureand also reflectsmetabolism. Theresulting apparentdiffusion coefficient istissue-specific and is ameasure of the mobilityof the water moleculesin the tissue.

h The principal advantageof MRI is the abilityto investigatetissue-dependentresponses to a magneticfield in multiple planeswithout the use ofionizing radiation.


Introduction to MRI


midsagittal line oriented along theanterior and posterior commissureor alternatively along the genu andsplenium of the corpus callosum.Coronal acquisitions should be ori-ented along the principle targetsof the examination, ie, in an 18- to30-degree angulation perpendicularto the long axis of the hippocampus toassess temporal lobe atrophy or mesio-temporal sclerosis, along the pituitarystalk in cases of sellar pathology, oralong the main axis of the optic nervesin orbital pathology. Sagittal slicesshould be oriented along the midlineof the corpus callosum. The appropri-ate sequences should follow predefinedand standardized protocols. However,during acquisition, changes may benecessary to follow unexpected pathol-ogies, and one should not rely onprotocols that may be hampered byinadequate slice thickness or lack ofDWI, GRE/T2*, or susceptibility-weighted imaging (SWI) sequences.

For standard acquisitions, the authorsrecommend DWI/ADC, T1-weighted,T2-weighted, and FLAIR sequences asa basic standard, with optional acquisi-tions of time-of-flight, SWI, and T1-weighted gadolinium sequences with aspatial resolution between 3 mm and5 mm (depending on field strength)with at least one high-resolution three-dimensional acquisition with isotropicvoxels (each side of the voxel equals thesame length, here 1.0 mm � 1.0 mm �1.0 mm) (Table 1-1).

A stroke protocol for MRI shouldencompass DWI, FLAIR, T2*/SWI, dy-namic susceptibility contrast perfu-sion, and time-of-flight or first-passgadolinium MR angiography (MRA).In cases of hemorrhage, additionalT1-weighted sequences may be help-ful to inform about the onset of thehemorrhage. Venous time-of-flight an-giograms and FLAIR and T1-weightedsequences with and without gadolin-

ium detect venous thrombosis; T1-and T2-weighted fat-suppressedsequences along the cervical arteriesdetect dissections of the arteries.Vascular malformations require time-resolved ultrafast angiograms of theaffected vessels with a temporal reso-lution greater than 1.0 second.

Specialized high-resolution proto-cols are also recommended whensearching for epileptogenic lesions inpatients with (focal) epilepsy syn-dromes, since conventional protocolsmay miss the abnormalities in approxi-mately 50% of cases.7 Six sequences areconsidered as essential, encompassingT1-weighted, high-resolution T1-weighted, T2/STIR, FLAIR in coronaland axial acquisitions, and GRE/SWI se-quences of 3 mm or more. In patientsin whom there is a concern for lepto-meningeal inflammation and infiltra-tion, standard routine protocols maybe complemented by postcontrastFLAIR. For differentiation betweeninflammatory pseudotumoral lesionsand CNS tumors, DWI, perfusion imag-ing, and MR spectroscopy may be ofadditional value (Figure 1-4).

Beyond lesion analysis, standardclinical routine MRI protocols embodya wealth of information about thefunctional state of the brain. DWI,for example, is now part of most rou-tine protocols. Diffusion is restrictedin tissue with a dense cellular matrixor in the presence of cytotoxic edemawith subsequent lowering of the ADC(Figure 1-5).

While DWI is frequently employedto detect the core of an ischemicstroke8 or a brain abscess,9 it may alsoaid in the differential diagnosis betweenseizure-related diffusion restriction andan ischemic stroke,10 in tumefactivemultiple sclerosis lesions,11 in gliomaand lymphoma,12 and in the prognos-tic assessment of coma13 and traumaticbrain injury.14,15



TABLE

1-1

RoutineMagneticReso

nance

Imaging

Pro

toco

ls

Sequen

ceDWI/ADC

T2-W

eighted

T1-W

eighted

FLAIR

SWITo

FCEMRA

Perfusion

CE

T1-W

eighted

FLAIR+CE

Mag

netic

resonan

cescreen

ing

Axial

Axial

Axial

Coronal

Axial

Optional

Sagittal

3-D

GRE

Stroke

Axial

Axial

Axial

Optional

Intracraniala

nd

extracranial

Axial

Axial

Axial

Trau

ma

Axial

Axial

Sagittal

Axial

Axial

Multiple

sclerosis

Axial

3mm

or3-D

Sagittal3-D

GRE

Sagittal3

-DSa

gittal3-D

GREan

dT1

-weighted

spin

echo

Epile

psy

Axial

T2/STIRax

ial/

coronal

3-D

GRE

FLAIR

axial/

coronal

Axial

Glio

ma

Axial

Axial

3-D

GRE

Axial

Optional

Sagittal3-D

GREan

dT1

-weigh

tedSE

Infections

Axial

Axial

Axial/sag

ittal

Axial

Sagittal3-D

GREan

dT1

-weigh

tedSE

Coronal

Neu

rode

gene

rative

disorders

Axial

Axial

3-D

GRE

Coronal

Axial

Axial

Optional:A

SLOptional

3-D=three-dimen

sion

al;A

DC=ap

parent

diffusioncoefficient;A

SL=arteria

lspinlabe

ling;

CE=contrast-enh

anced;

DWI=

diffusion-weigh

tedim

aging;

FLAIR

=flu

id-atten

uatedinversion

recovery;G

RE=grad

ient

recalledecho

;MRA

=mag

netic

resona

ncean

giog

raph

y;SE

=spin

echo

;STIR=shorttauinversionrecovery;S

WI=

suscep

tibility-w

eigh

tedim

aging;

ToF=tim

eof

fligh

t.


Introduction to MRI


MRI may aid also in the differenti-ation between vasogenic tumoraledema and tumor infiltration,16 intreatment selection of patientsexperiencing wake-up strokes withundefined onset,17 and in assess-ment of various CSF pathologies,such as meningitis, meningeal carci-nomatosis, stroke, and hemorrhage18

(Figure 1-6).SWI is used to assess disorders

associated with hemoglobin break-down (eg, parenchymal hemorrhage,cerebral amyloid angiopathy, cavernousmalformations, subarachnoid hemor-

rhage), calcifications, hemorrhagic con-tusions in traumatic brain injury, oraltered blood oxygenation due to strokeor status epilepticus (Figure 1-7).19Y20

Perfusion imaging in acute strokehas been shown to be a good predic-tor of the ischemic penumbra21 butmay also aid in the differentiationbetween ischemic and postictalhypoperfusion22 or in migraine attackswith atypical aura symptoms.23 Typicalperfusion findings in different clinicalscenarios can be seen in Figure 1-8. Inneuro-oncology, cell density andprominent vascularization reflect

FIGURE 1-4 Signal characteristics of a tumefactive demyelinating lesion. A, Faint fluid-attenuated inversion recovery(FLAIR) hyperintensity; B, C, diffusion restriction and decreased apparent diffusion coefficient (ADC) imaging.D, Cerebral blood flow maps indicate hyperperfusion. E, Follow-up MRI (ADC map) after 24 hours indicateslesion expansion and ADC signal drop following acute demyelination.

FIGURE 1-5 Patterns of diffusion-weighted imaging (DWI) restrictions in various neurologic disorders. A, Right middlecerebral artery (MCA) territory infarction; B, Creutzfeldt-Jakob disease (cortical, basal ganglia); C,symmetrically cortical in hypoxic-ischemic encephalopathy (HIE); D, cortical mixed cytotoxic and vasogenicpattern in mitochondrial encephalomyopathy, lactic acidosis, and strokelike episodes syndrome (MELAS).



heterogeneous patterns of malignancythan can be prognostically assessed bycombinations of DWI and perfusionimaging24 to differentiate tumor pro-gression and therapy-related effects.25

GENERAL CONSIDERATIONSFOR IMAGE INTERPRETATIONThe way MRI is used in clinical prac-tice strongly depends on the clinicalneeds and context of application,whether as an emergency MRI within

10 to 15 minutes because of a strokeor in an otherwise uncooperativepatient, or as an elective structural orfunctional imaging study, which maylast around 1 hour and requires ex-tensive postprocessing procedures af-terward. The underlying disorder maybe misdiagnosed if, for example, anuncooperative patient is studied witha rigidly applied stroke protocol in anemergency setting. Imaging findingsmay be so subtle that only with the

FIGURE 1-6 Coronal fluid-attenuated inversion recovery (FLAIR) imaging indicating three characteristic patterns ofhyperintensity. A, Bilateral hippocampal and parahippocampal swelling in tick-borne encephalitis; B,bithalamic and diencephalic edema due to venous thrombosis of the internal cerebral veins; C, dirty CSFsign in a patient with subarachnoid hemorrhage (SAH). Note the hyperintensity due to expansion of bloodinto the sulcal space.

FIGURE 1-7 Advanced MRI applications for ischemic stroke. A, Axial susceptibility-weighted imaging (SWI) indicateslocation and thrombus length in the left middle cerebral artery (arrow); B, digital subtraction angiographyindicates vessel occlusion of the left internal carotid artery (arrow); C, SWI indicates prominent veins due to

increased oxygen extraction (paramagnetic effect) (arrow); D, perfusion imaging ([time-to-maximum; max delay] longer than6 s indicated in red) (arrow) corresponds to the area with prominent vessels on SWI.


Introduction to MRI


appropriate clinical information willa lesion be found or correctly inter-preted (Figure 1-9).

To interpret images correctly, read-ing and analysis of the MRI shouldincorporate a systematic approach.This may encompass a structuredsequence of reading, first inspectingthe DWI/ADC sequences (especially incases where the main findings can beexpected from facilitated or restricted

diffusion), then switching to T2/FLAIRsequences and T1-weighted se-quences with and without contrast toget a first overview of suspectedpathologies. In a second step, theremaining sequences should be ana-lyzed in the context of the suspectedpathology, eg, perfusion imaging, SWI,and vascular sequences if a vasculardisorder is suspected. A systematicapproach should consider fixed reading

FIGURE 1-9 Imaging of a 76-year-old man with right-sided hemichoreic movements for several weeks. A, Axial fluid-attenuatedinversion recovery (FLAIR) sequence is unremarkable. B, Axial cerebral blood volume map shows a slight elevationin cerebral blood volume in the left lentiform nucleus (arrow). C, Axial three-dimensional time-of-flight magnetic

resonance angiogram (MRA) shows diminished signal of the left internal carotid (arrow) and middle cerebral artery suggesting ahigh-grade proximal internal carotid artery stenosis.

FIGURE 1-8 Perfusion imaging indicating abnormal cerebral blood flow maps in different neurologic disorders. A, Patientwith postictal hypoperfusion in the left temporal lobe presenting with aphasia. B, Patient with hypoperfusion(exceeding the vascular territories) in the right temporal and occipital lobe during a migraine attack with

headache. C, Patient with clinical presentation of frontotemporal dementia and a symmetric pattern of reduced cerebralblood flow in the frontal lobes. Arrows indicate brain areas with regional perfusion decrease.



procedures, for example, from insideto outside or vice versa, from graymatter to white matter and CSF, orfrom suspected pathology to generalinspection on a slice-by-slice basis. Itcannot be emphasized enough thatthe peripheral parts of the imagesneed to beinspected carefully; mostpathologies present in the peripherymay be missed if those areas arenot examined carefully by the imagereader. If possible, images shouldbe read twice, either by the neurora-diologist and the neurologist or bytwo specialized readers, to enable con-sistency checks of the suspected find-ings, with the final goal to identifyand interpret structural lesions. If alesion is detected, it should first beclassified by broad category, eg, tumorallesion, vascular pathology, infection,or degeneration. Depending on theunderlying pathology, typical presen-tations and imaging features may betaken into account:

& Cerebral tumors: Slowly progressivefocal clinical deficit, cognitive

decline, or epileptic seizures. MRIindicates intraaxial or extraaxialpathology, a space-occupying effect,gadolinium enhancement.vasogenic edema, and midline shift,

& Vascular: Sudden onset of aneurologic deficit. MRI informationmay include vascular territory andDWI lesion with ADC reduction.

& Infections: Headache, fever, focalneurologic deficit, or increasedcells in CSF. MRI findings includeswelling of cerebral sulci ormeningeal enhancement.

& Degenerative diseases: Cognitivedecline, epileptic seizures. MRI mayindicate focal or generalized atrophy,no gadolinium enhancement, andno focal lesions.

If no lesion is found, one should care-fully reanalyze the images and take intoconsideration that a lesion may havebeen overlooked. A lesion may beoverlooked because of an insufficientspatial resolution or insufficient contrastto the surrounding tissue (Figure 1-10)

KEY POINTS

h A systematic approachshould consider fixedreading procedures, forexample, from inside tooutside or vice versa,from gray matter towhite matter and CSF,or from suspectedpathology to generalinspection on aslice-by-slice basis.

h If a lesion is detected, itshould first be classifiedby broad category, eg,tumoral lesion, vascularpathology, infection,or degeneration.

FIGURE 1-10 Focal cortical dysplasia type IIB. A, Coronal two-dimensional fluid-attenuated inversion recovery (FLAIR)turbo spin echo sequence with a slice thickness of 2 mm; B, coronal reformation of a three-dimensionalFLAIR sampling perfection with application-optimized contrasts using different flip angle evolutions (SPACE)

sequence; C, sagittal three-dimensional FLAIR SPACE sequence (1 mm � 1 mm � 1 mm voxel). Despite a highersignal to noise ratio, the funnel-shaped hyperintensity is hardly visible on the two-dimensional sequence (A, arrow).Multiplanar coronal reformation (B) of the three-dimensional FLAIR sequence helps locate the lesion, which is typicallyfound in the depth of a sulcus (B, C, arrows).


Introduction to MRI


or because of misleading clinical infor-mation or a lack of clinical informa-tion. In some instances, a lesion maynot yet be visible, eg, during an earlymanifestation of an acute spinal in-farction (Figure 1-11) or in the caseof a lesion that is responsible for theclinical symptoms but hidden amongmany others.

Although lesion detectability is de-pendent on the contrast to noise ratio,ie, the signal to noise ratio of the lesionversus the signal to noise ratio of thesurrounding tissue, in most cases, for a

lesion to be detected, it must be at least2 voxels (volume element defined bythe slice thickness and the in-planeresolution) in size. If a lesion is lessthan 2 voxels in size, it may easily bemissed because of partial volume effects.

Two-dimensional sequences havea higher in-plane resolution and ahigher signal to noise ratio than three-dimensional sequences26; however,they rarely cover the entire brain witha slice thickness of 1 mm to 2 mm.Three-dimensional sequences are gen-erally acquired in sagittal orientation

FIGURE 1-11 A 31-year-old man with sudden onset of pain and paraplegia below the level ofT10. A, Initial MRI with T2-weighted fast spin echo and sagittal diffusion-weightedimaging (not shown) was unremarkable. B, Follow-up axial imaging3days later shows

a left-sided posterolateral spinal infarct. The insets in panel B are the axial images at the positionsshown by the thick white lines in the sagittal T2-weighted fast spin echo (left) and diffusion-weighted(right) images. The thin white lines represent the range that is covered by the axial sequences.



and can be reformatted in any planeusing a multiplanar reformation tool.This multiplanar reformation helps tofind subtle (epileptogenic) lesions(Figure 1-10). On the other hand,high in-plane resolutionmay be neededto show the details of a lesion andmakea specific diagnosis (Figure 1-12).

Image interpretation should alwaysdepend on the combination and inte-gration of imaging features and clinicalinformation. While the signal intensitymay be a specific characteristic for asuspected lesion (eg, the high DWIsignal of epidermoid cysts, abscesses,and infarctions), a combination ofseveral imaging parameters are ofcrucial importance to narrow thedifferential diagnosis. A checklist ofcharacteristic imaging features to beanalyzed ahead of image interpreta-tion are summarized in Table 1-2.

Since location is a highly relevantparameter for diagnosis, a strongeffort should be made to categorize alesion with respect to its topography.Try to exactly locate a lesion withrespect to: the spaces covering the

brain (epidural, subdural, subarach-noid space), cortex, white matter,ventricular ependymal, and ventricularspace. The location of a lesion isappropriately assessed with the lesioncut perpendicular to its surface andconsidering the effect on the surround-ings (Figure 1-13). In most cases, anappropriate reformation of a three-dimensional sequence (eg, FLAIR,magnetization-prepared rapid ac-quisition gradient-echo imaging[MPRAGE]) is considered helpful.

The signal intensity of a lesion mayprovide information about its compo-sition. A lesion is classified as hypoin-tense, isointense, or hyperintense ascompared to the gray matter if nototherwise stated. T1-hyperintense le-sions consist of fat, blood in themethemoglobin stage, or proteina-ceous fluid, or show contrast enhance-ment. Blood vessels typically are voidof any signal (flow void); they may,however, show a paradoxical enhance-ment on noncontrast T1-weightedGRE sequences.

KEY POINTS

h Image interpretationshould always dependon the combination andintegration of imagingfeatures and clinicalinformation.

h Since location is a highlyrelevant parameter fordiagnosis, a strongeffort should be madeto categorize a lesionwith respect to itstopography.

h The signal intensity ofa lesion may provideinformation about itscomposition. Alesion is classified ashypointense, isointense,or hyperintense ascompared to the graymatter if nototherwise stated.

h T1-hyperintense lesionsconsist of fat, blood inthe methemoglobinstage, or proteinaceousfluid, or show contrastenhancement.

FIGURE 1-12 High-resolution T2-weighted short tau inversion recovery (STIR) sequence(0.45 mm � 0.45 mm � 2 mm) shows a nodule within a cyst (A, arrow)suggestive of the scolex of the cysticercus. Ringlike contrast enhancement ofthe lesion supports the diagnosis (B).


Introduction to MRI


The signal intensity of lesions thatare T1 isointense, T1 hypointense, andT2 hypointense, isointense, or hyper-intense is related to the cellularity ofthese lesions. Lesions with a highcellularity (high nucleus to cytoplasmratio) show a relatively high signal onT1-weighted images and a relativelylow signal on T2-weighted images(Figure 1-14).

DWI provides information aboutthe mobility of free water protonswithin a lesion, complementing SWIand T2-weighted sequences, whichreveal whether a lesion contains calci-fication or blood degradation prod-ucts. Sometimes additional applicationof CT may be necessary to confirmthe diagnosis of a calcification or acavernous malformation or to showthe osseous structures more clearly(Figure 1-15).

Interpretation of pathologic imag-ing features require a fundamentalknowledge of normal anatomy andsignal and their variants. Careful con-sideration of the localization of le-sions, space-occupying effects, signalabnormalities of T1 and T2, andcontrast enhancements help deter-mine whether a correct diagnosis ismade based on MRI.

CONCLUSION

MRI produces images that representsignal intensities of tissue that aredependent on relaxation time and spa-tial resolution. Pathologic structures areidentified by specific tissue behaviormainly related to different relaxationtimes. Combined with knowledge ofclinical symptoms and signs, and recog-nition of suspected localization andetiology, careful assessment and inter-pretation of MRI scans is an essentialdiagnostic skill.

KEY POINT

h MRI produces imagesthat represent signalintensities of tissue thatare dependent onrelaxation time andspatial resolution.

TABLE 1-2 Characteristic ImagingFeatures to BeAnalyzed Ahead ofImage Interpretation

b Location

Intraaxial, extraaxial(epidural, subdural, CSF space)

Lobe, gyrus, ventricular system

Gray matter, white matter(central, peripheral,U fiber involvement)

b Morphology

Exophytic component

Cystic component

Encapsulated process

Central necrosis

Calcifications

Hemorrhage

Perifocal edema

b Signal Intensity

Spin density, T1 relaxation,T2 relaxation

Susceptibility

Diffusion (water mobility)

Flow (flow void)

b Delineation

b Contrast Enhancement

Blood-brainbarrierdisturbance

Vessels

b Effect on Environment

Space occupying

Not space occupying

Space giving

Bone destruction

Bone remodeling

b Growth Pattern

Infiltrative

Multiplicity

Metastases



FIGURE 1-13 Ganglioglioma with remodeling of the adjacent bone. A, Axial fluid-attenuatedinversion recovery (FLAIR) MRI; B, coronal T2-weighted fast spin echo sequence;C, CT shows a tiny cortical partially calcified lesion with exophytic growth/bonyremodeling (A, B, C, arrows).

FIGURE 1-14 T2 lesions with relative hypointensity suggestive for highly cellular lesions suchas lymphoma (A, arrow), vestibular schwannoma, or, as in this case,meningioma (B, arrow) and metastasis (C, arrow).

FIGURE 1-15 A 12-year-old girl with a teratoma in the pineal region. A clue to the diagnosis isthe mixture of fat and calcified elements on T2-weighted (A) and T1-weighted(B) images. Fat is hyperintense on T1-weighted and T2-weighted images;calcification may show different signal intensities. A complementary CT scan (C )helps to differentiate fat as hypodense (blue arrow) and calcification as stronglyhyperdense (open arrow).


Introduction to MRI


REFERENCES1. Horowitz AL. MRI physics for physicians.

New York, NY: Springer Verlag, 1989.

2. McRobbie DW, Moore EA, Graves MJ,Prince MR. MRI from picture to proton.Cambridge, UK: Cambridge UniversityPress, 2007.

3. Schild HH. MRI made easy. Berlin, Germany:Nationales Druckhaus & Schering, 1990.

4. Jung BA, Weigel M. Spin echo magneticresonance imaging. J Magn ResonImaging 2013;37(4):805Y817. doi:10.1002/jmri.24068.

5. Markl M, Leupold J. Gradient echo imaging.J Magn Reson Imaging 2012;35(6):1274Y1289. doi:10.1002/jmri.23638.

6. Johansen-Berg H, Behrens TEJ. DiffusionMRI: from quantitative measurement to invivo neuroanatomy. Cambridge, MA:Academic Press, 2013.

7. Wellmer J, Quesada CM, Rothe L, et al.Proposal for a magnetic resonanceimaging protocol for the detection ofepileptogenic lesions at early outpatientstages. Epilepsia 2013;54(11):1977Y1987.doi:10.1111/epi.12375.

8. Albers GW, Thijs VN, Wechsler L, et al.Magnetic resonance imaging profilespredict clinical response to early reperfusion:the diffusion and perfusion imagingevaluation for understanding strokeevolution (DEFUSE) study. Ann Neurol2006;60(5):508Y517. doi:10.1002/ana.20976.

9. Desprechins B, Stadnik T, Koerts G, et al.Use of diffusion-weighted MR imagingin differential diagnosis betweenintracerebral necrotic tumors and cerebralabscesses. AJNR Am J Neuroradiol 1999;20(7):1252Y1257.

10. Szabo K, Poepel A, Pohlmann-Eden B,et al. Diffusion-weighted and perfusionMRI demonstrates parenchymal changes incomplex partial status epilepticus. Brain2005;128(pt 6):1369Y1376. doi:10.1093/brain/awh454.

11. Malhotra HS, Jain KK, Agarwal A, et al.Characterization of tumefactivedemyelinating lesions using MR imagingand in-vivo proton MR spectroscopy.Mult Scler 2009;15(2):193Y203.doi:10.1177/1352458508097922.

12. Yamasaki F, Kurisu K, Satoh K, et al.Apparent diffusion coefficient of humanbrain tumors at MR imaging. Radiology2005;235(3):985Y991. doi:10.1148/radiol.2353031338.

13. Choi SP, Park KN, Park HK, et al.Diffusion-weighted magnetic resonanceimaging for predicting the clinical outcomeof comatose survivors after cardiac arrest:

a cohort study. Crit Care 2010;14(1):R17.doi:10.1186/cc8874.

14. Huisman TA, Sorensen AG, Hergan K,et al. Diffusion-weighted imaging for theevaluation of diffuse axonal injury inclosed head injury. J Comput Assist Tomogr2003;27(1):5Y11.

15. Karaarslan E, Arslan A. Diffusionweighted MR imaging in non-infarctlesions of the brain. Eur J Radiol2008;65(3):402Y416. doi:10.1016/j.ejrad.2007.04.023.

16. Artzi M, Bokstein F, Blumenthal DT, et al.Differentiation between vasogenic-edemaversus tumor-infiltrative area in patientswith glioblastoma during bevacizumabtherapy: a longitudinal MRI study. Eur JRadiol 2014;83(7):1250Y1256. doi:10.1016/j.ejrad.2014.03.026.

17. Thomalla G, Cheng B, Ebinger M, et al.DWI-FLAIR mismatch for the identificationof patients with acute ischaemic strokewithin 4.5 h of symptom onset (PRE-FLAIR):a multicentre observational study. LancetNeurol 2011;10(11):978Y986. doi:10.1016/S1474-4422(11)70192-2.

18. Stuckey SL, Goh TD, Heffernan T, Rowan D.Hyperintensity in the subarachnoid spaceon FLAIR MRI. AJR Am J Roentgenol2007;189(4):913Y921. doi:10.2214/AJR.07.2424.

19. Aellen J, Abela E, Buerki SE, et al.Focal hemodynamic patterns of statusepilepticus detected by susceptibilityweighted imaging (SWI). Eur Radiol2014;24(11):2980Y2988. doi:10.1007/s00330-014-3284-9.

20. Ong BC, Stuckey SL. Susceptibilityweighted imaging: a pictorial review.J Med Imaging Radiat Oncol2010;54(5):435Y449. doi:10.1111/j.1754-9485.2010.02208.x.

21. Olivot JM, Mlynash M, Thijs VN, et al.Geography, structure, and evolutionof diffusion and perfusion lesionsin Diffusion and perfusion imagingEvaluation For Understanding StrokeEvolution (DEFUSE). Stroke 2009;40(10):3245Y3251. doi:10.1161/STROKEAHA.109.558635.

22. Hauf M, Slotboom J, Nirkko A, et al.Cortical regional hyperperfusion innonconvulsive status epilepticusmeasured by dynamic brainperfusion CT. AJNR Am JNeuroradiol 2009;30(4):693Y698.doi:10.3174/ajnr.A1456.

23. Forster A, Wenz H, Kerl HU, et al. Perfusionpatterns in migraine with aura. Cephalalgia2014;34(11):870Y876. doi:10.1177/0333102414523339.



24. Deike K, Wiestler B, Graf M, et al.Prognostic value of combined visualizationof MR diffusion and perfusion mapsin glioblastoma. J Neurooncol2016;126(3):463Y472. doi:10.1007/s11060-015-1982-z.

25. Farid N, Almeida-Freitas DB, White NS,et al. Combining diffusion andperfusion differentiates tumor from

bevacizumab-related imaging abnormality(bria). J Neurooncol 2014;120(3):539Y546.doi:10.1007/s11060-014-1583-2.

26. Tschampa HJ, Urbach H, Malter M, et al.Magnetic resonance imaging of focalcortical dysplasia: comparison of 3D and 2Dfluid attenuated inversion recoverysequences at 3T. Epilepsy Res 2015;116:8Y14.doi:10.1016/j.eplepsyres.2015.07.004.


Introduction to MRI


introductiontomagnetic resonance imaging for basel ... · standing of mri by introducing the basics...

Documents