Diffusion-weighted magnetic resonance (MR) imaging provides image contrast thatis different from that provided by conventional MR techniques. It is particularlysensitive for detection of acute ischemic stroke and differentiation of acute strokefrom other processes that manifest with sudden neurologic deficits. Diffusion-weighted MR imaging also provides adjunctive information for other cerebral dis-eases including neoplasms, intracranial infections, traumatic brain injury, and de-myelinating processes. Because stroke is common and in the differential diagnosis ofmost acute neurologic events, diffusion-weighted MR imaging should be consideredan essential sequence, and its use in most brain MR studies is recommended.
Diffusion-weighted (DW) magnetic resonance (MR) imaging provides potentially uniqueinformation on the viability of brain tissue. It provides image contrast that is dependenton the molecular motion of water, which may be substantially altered by disease. Themethod was introduced into clinical practice in the middle 1990s, but because of itsdemanding MR engineering requirements—primarily high-performance magnetic fieldgradients—it has only recently undergone widespread dissemination. The primary appli-cation of DW MR imaging has been in brain imaging, mainly because of its exquisitesensitivity to ischemic stroke—a common condition that appears in the differential diag-nosis in virtually all patients who present with a neurologic complaint.
Because DW MR imaging uses fast (echo-planar) imaging technology, it is highlyresistant to patient motion, and imaging time ranges from a few seconds to 2 minutes. Asa consequence, DW MR imaging has assumed an essential role in the detection of acutebrain infarction and in the differentiation of acute infarction from other disease processes.DW MR imaging is also assuming an increasingly important role in the evaluation of manyother intracranial disease processes.
BASIC CONCEPTS OF DW MR IMAGING
In this section, the basic concepts involved in DW MR imaging will be briefly reviewed. Formore detailed descriptions of the physics of DW imaging, a number of excellent reviews(1–5) are available. Stejskal and Tanner (6) provided an early description of a DW sequencein 1965. They used a spin-echo T2-weighted pulse sequence with two extra gradient pulsesthat were equal in magnitude and opposite in direction. This sequence enabled themeasurement of net water movement in one direction at a time. To measure the rate ofmovement along one direction, for example the x direction, these two extra gradients areequal in magnitude but opposite in direction for all points at the same x location.However, the strength of these two balanced gradients increases along the x direction.Therefore, if a voxel of tissue contains water that has no net movement in the x direction,the two balanced gradients cancel each other out. The resultant signal intensity of thatvoxel is equal to its signal intensity on an image obtained with the same sequence withoutthe DW gradients. However, if water molecules have a net movement in the x direction(eg, due to diffusion), they are subjected to the first gradient pulse at one x location andthe second pulse at a different x location. The two gradients are no longer equal inmagnitude and no longer cancel. The difference in gradient pulse magnitude is propor-tional to the net displacement in the x direction that occurs between the two gradientpulses, and faster-moving water protons undergo a larger net dephasing. The resultantsignal intensity of a voxel of tissue containing moving protons is equal to its signal
State of the Art
331
intensity on a T2-weighted image de-creased by an amount related to the rateof diffusion.
The signal intensity (SI) of a voxel oftissue is calculated as follows:
SI 5 SI0 3 exp ~2b 3 D!, (1)
where SI0 is the signal intensity on theT2-weighted (or b 5 0 sec/mm2) image,the diffusion sensitivity factor b is equalto g2G2d2(D 2 d/3), and D is the diffusioncoefficient. g is the gyromagnetic ratio; Gis the magnitude of, d the width of, and Dthe time between the two balanced DWgradient pulses.
According to Fick’s law, true diffusionis the net movement of molecules due toa concentration gradient. With MR imag-ing, molecular motion due to concentra-tion gradients cannot be differentiatedfrom molecular motion due to pressuregradients, thermal gradients, or ionic in-teractions. Also, with MR imaging we donot correct for the volume fraction avail-able or the increases in distance traveleddue to tortuous pathways. Therefore,when measuring molecular motion withDW imaging, only the apparent diffusioncoefficient (ADC) can be calculated. Thesignal intensity of a DW image is bestexpressed as
SI 5 SI0 3 exp ~2b 3 ADC!. (2)
With the development of high-perfor-mance gradients, DW imaging can beperformed with an echo-planar spin-echo T2-weighted sequence. With theoriginal spin-echo T2-weighted sequence,even minor bulk patient motion wasenough to obscure the much smaller mo-lecular motion of diffusion. The substitu-tion of an echo-planar spin-echo T2-weighted sequence markedly decreasedimaging time and motion artifacts andincreased sensitivity to signal changesdue to molecular motion. As a result,the DW sequence became clinically fea-sible to perform. Other methods of per-forming DW MR imaging without echo-planar gradients have also been devel-oped. These include DW sequences basedon a single-shot gradient and spin-echoor single-shot fast spin-echo techniques(7,8). “Line-scan” DW and spiral DWsequences have also been developed (9–12).
In the brain, apparent diffusion is notisotropic (the same in all directions); it isanisotropic (varies in different direc-tions), particularly in white matter. Thecause of the anisotropic nature of whitematter is not completely understood, butincreasing anisotropy has also been
noted in the developing brain before T1-and T2-weighted imaging or histologicevidence of myelination becomes evi-dent (13,14). It is likely that in additionto axonal direction and myelination,other physiologic processes, such as ax-olemmelic flow, extracellular bulk flow,capillary blood flow, and intracellularstreaming, may contribute to white mat-ter anisotropy. The anisotropic nature ofdiffusion in the brain can be appreciatedby comparing images obtained with DWgradients applied in three orthogonal di-rections (Fig 1). In each of the images, thesignal intensity is equal to the signal in-tensity on echo-planar T2-weighted im-ages decreased by an amount related tothe rate of diffusion in the direction ofthe applied gradients. Images obtainedwith gradient pulses applied in one direc-tion at a time are combined to create DWimages or ADC maps. The ADC is actuallya tensor quantity or a matrix:
ADC 5 F ADCxx ADCxy ADCxz
ADCyx ADCyy ADCyz
ADCzx ADCzy ADCzz
G. (3)
The diagonal elements of this matrixcan be combined to give informationabout the magnitude of the apparent dif-fusion: (ADCxx 1 ADCyy 1 ADCzz)/3.
The off-diagonal elements provide in-formation about the interactions be-tween the x, y, and z directions. For ex-ample, ADCyx gives information aboutthe correlation between displacements inthe x and y directions (4). Images display-
ing the magnitude of the ADC are used inclinical practice.
DW gradient pulses are applied in onedirection at a time. The resultant imagehas information about both the directionand the magnitude of the ADC (Fig 1). Tocreate an image that is related only tothe magnitude of the ADC, at least threeof these images must be combined. Thesimplest method is to multiply the threeimages created with the DW gradientpulses applied in three orthogonal direc-tions. The cube root of this product is theDW image (Fig 2). It is important to un-derstand that the DW image has T2-weighted contrast as well as contrast dueto differences in ADC. To remove the T2-weighted contrast, the DW image can bedivided by the echo-planar spin-echo T2-weighted (or b 5 0 sec/mm2) image togive an “exponential image” (Fig 3). Al-ternatively, an ADC map, which is animage whose signal intensity is equal tothe magnitude of the ADC, can be cre-ated (Fig 4).
Instead of obtaining images with b 5 0sec/mm2 and with b 5 1,000 sec/mm2
and solving for ADC using Equation (2),one usually determines the ADC graphi-cally. This is accomplished by obtainingtwo image sets, one with a very low butnonzero b value and one with b 5 1,000sec/mm2. By plotting the natural loga-rithm of the signal intensity versus b forthese two b values, the ADC can be de-termined from the slope of this line.
For our clinical studies, the DW image,
Figure 1. Anisotropic nature of diffusion in the brain. Transverse DW MR images (b 5 1,000sec/mm2; effective gradient, 14 mT/m; repetition time, 7,500 msec; minimum echo time; matrix,128 3 128; field of view, 200 3 200 mm; section thickness, 6 mm with 1-mm gap) with thediffusion gradients applied along the x (Gx, left), y (Gy, middle), and z (Gz, right) axes demon-strate anisotropy. The signal intensity decreases when the white matter tracts run in the samedirection as the DW gradient because water protons move preferentially in this direction. Notethat the corpus callosum (arrow on left image) is hypointense when the gradient is applied in thex (right-to-left) direction, the frontal and posterior white matter (arrowheads) are hypointensewhen the gradient is applied in the y (anterior-to-posterior) direction, and the corticospinal tracts(arrow on right image) are hypointense when the gradient is applied in the z (superior-to-inferior)direction.
332 z Radiology z November 2000 Schaefer et al
exponential image, ADC map, and echo-planar spin-echo T2-weighted images areroutinely available for review (Fig 4). Be-cause the ADC values of gray and whitematter are similar, typically there is nocontrast between gray and white matteron the exponential image or ADC map.The contrast between gray and whitematter seen on the DW image is due toT2-weighted contrast. This residual T2component on the DW image makes itimportant to view either the exponentialimage or ADC map in conjunction withthe DW image. In lesions such as acutestroke, the T2-weighted and DW effectsboth cause increased signal intensity onthe DW image. Therefore, we have foundthat we identify regions of decreased dif-fusion best on DW images. The exponen-tial image and ADC maps are used toexclude “T2 shine through” as the causeof increased signal intensity on DW im-ages. The exponential image and ADCmap are useful for detecting areas of in-creased diffusion that may be masked byT2 effects on the DW image.
CLINICAL APPLICATIONS
The Table provides a summary of DWand ADC imaging findings, as well as thecharacteristic ADC and causes, for a vari-ety of disease entities.
Ischemic Stroke
Theory of restricted diffusion in acutestroke.—Within minutes after the onsetof ischemia, a profound restriction in wa-ter diffusion occurs in affected brain tis-sue (15–18). The biophysical basis of thischange is not completely clear. Onelikely important contributor is cytotoxicedema. Cytotoxic edema induced withacute hyponatremic encephalopathy (with-out ischemia) is associated with restricteddiffusion (19). Furthermore, when de-creased ADCs were present in early isch-emia in rat brain tissue, there was a re-duction in Na1/K1 adenosine triphos-phatase activity without a significant in-crease in brain water (16). In addition,ouabain, an inhibitor of Na1/K1 adeno-sine triphosphatase, was associated witha reduction in ADC in rat cortex (20).These findings have led to the predomi-nant theory for the restriction of waterdiffusion in stroke: Ischemia causes dis-ruption of energy metabolism, leading tofailure of the Na1/K1 adenosine triphos-phatase pump and other ionic pumps.This leads to loss of ionic gradients anda net translocation of water from theextracellular to the intracellular compart-
Figure 2. Calculation of signal intensity on an isotropic DW image (b 5 1,000 sec/mm2;effective gradient, 14 mT/m; repetition time, 7,500 msec; minimum echo time; matrix, 128 3128; field of view, 400 3 200 mm; section thickness, 6 mm with 1-mm gap). The signal intensitiesof the three transverse images (Gx, Gy, and Gz), each with a diffusion gradient applied in one ofthree orthogonal directions, are multiplied together. Here the DW gradients were applied alongthe x, y, and z axes. The signal intensity of the isotropic DW image (bottom) is essentially thecube root of the signal intensities of these three images multiplied together. Note that bothT2-weighted contrast and the rate of diffusion contribute to the signal intensity of the isotropicDW image.
Figure 3. Removal of T2-weighted contrast. To remove the T2-weighted contrast in the isotropictransverse DW image (b 5 1,000 sec/mm2; effective gradient, 14 mT/m; repetition time, 7,500msec; minimum echo time; matrix, 128 3 128; field of view, 200 3 200 mm, section thickness,6 mm with 1-mm gap), the transverse DW image (DWI) is divided by the transverse echo-planarspin-echo T2-weighted (EP SET2) image. The resultant image is called the exponential imagebecause its signal intensity is exponentially related to the ADC.
Volume 217 z Number 2 Diffusion-weighted MR Imaging of the Brain z 333
ment, where water mobility is relativelymore restricted.
There are additional factors. With cel-lular swelling, there is a reduction in thevolume of extracellular space (21). A de-crease in the diffusion of low-molecular-weight tracer molecules has been demon-strated in animal models (22,23), whichsuggests that the increased tortuosity ofextracellular space pathways contributesto restricted diffusion in acute ischemia.Furthermore, there are substantial reduc-tions in ADCs in intracellular metabolitesin ischemic rat brain (24–26). Proposedexplanations are increased intracellularviscosity due to dissociation of microtu-bules and fragmentation of other cellularcomponents or increased tortuosity ofthe intracellular space and decreased cy-toplasmic mobility. It is worth bearing inmind that the normal steady-state func-tion of these structures requires energyand uses adenosine triphosphate. Otherfactors such as temperature (27,28) andcell membrane permeability (29,30) playa minor role in explaining the reductionin ADC in acutely ischemic tissue.
Time course of lesion evolution in acutestroke.—In animals, restricted diffusionassociated with acute ischemia has beendetected as early as 10 minutes to 2 hoursafter vascular occlusion (17,18,31–35).
The ADCs measured at these times areapproximately 16%–68% below those ofnormal tissue. In animals, ADCs pseudo-normalize (ie, are similar to those of nor-mal brain tissue, but the tissue is in-farcted) at approximately 2 days and areelevated thereafter.
In adult humans, the time course is
more prolonged (Fig 5) (36–39). We haveobserved restricted diffusion associatedwith acute ischemia 30 minutes after awitnessed ictus. The ADC continues todecrease and is most reduced at 8–32hours. The ADC remains markedly re-duced for 3–5 days. This decreased diffu-sion is markedly hyperintense on DW
Figure 4. Creation of an ADC map. One method of creating an ADC map is to mathematicallymanipulate the exponential (Exp.) image. The appearances on the transverse DW image (DWI),exponential image, and ADC map, as well as the corresponding mathematic expressions for theirsignal intensities, are shown. Image parameters are b 5 1,000 sec/mm2; effective gradient, 14mT/m; repetition time, 7,500 msec; minimum echo time; matrix, 128 3 128; field of view, 200 3200 mm; section thickness, 6 mm with 1-mm gap. SI 5 signal intensity, SIo 5 signal intensity onT2-weighted image.
DW MR Imaging Characteristics of Various Disease Entities
Disease
MR Signal Intensity
ADC CauseDW Image ADC Image
Acute stroke High Low Restricted Cytotoxic edemaChronic stroke Variable High Elevated GliosisHypertensive encephalopathy Variable High Elevated Vasogenic edemaCyclosporin toxicity Variable High Elevated Vasogenic edemaHyperperfusion after carotid endarterectomy Variable High Elevated Vasogenic edemaHIV encephalopathy Variable High Elevated Vasogenic edemaIntraaxial mass
Necrotic center Variable High Elevated Increased free waterSolid tumor Variable Variable Variable Depends on cellularity
Arachnoid cyst Low High Elevated Free waterEpidermoid mass High Low* Restricted* Cellular tumorPyogenic infection High Low Restricted ViscosityHerpes encephalitis High Low Restricted Cytotoxic edemaCreutzfeldt-Jakob syndrome High Low Restricted UnknownDiffuse axonal injury
Majority of cases High Low Restricted Cytotoxic edemaMinority of cases Variable High Elevated Vasogenic edema
HemorrhageOxyhemoglobin High Low Restricted IntracellularDeoxyhemoglobin Low Unknown† Unknown† Unknown†
Extracellular methemoglobin High High Elevated ExtracellularHemosiderin Low Unknown† Unknown† Unknown†
Multiple sclerosisMost acute lesions Variable High Elevated Vasogenic edemaA few acute lesions High Low Restricted UnknownChronic lesions Variable High Elevated Gliosis, neuronal loss
* Relative to that of cerebrospinal fluid (CSF).† The ADC usually cannot be calculated.
334 z Radiology z November 2000 Schaefer et al
images (which are generated with a com-bination of T2-weighted and DW imag-ing) and hypointense on ADC images.The ADC returns to baseline at 1–4weeks. This most likely reflects persistenceof cytotoxic edema (associated with de-creased diffusion) and development of va-sogenic edema and cell membrane disrup-tion, leading to increased extracellularwater (associated with increased diffusion).At this point, an infarction is usuallymildly hyperintense due to the T2 compo-nent on the DW images and is isointenseon the ADC images. Thereafter, diffusion iselevated as a result of continued increase inextracellular water, tissue cavitation, andgliosis. This elevated diffusion is character-ized by slight hypointensity, isointensity,or hyperintensity on the DW images (de-pending on the strength of the T2 anddiffusion components) and increased sig-nal intensity on ADC maps.
The time course does not always con-form to the aforementioned outline.With early reperfusion, pseudonormal-ization (return to baseline of the ADCreduction associated with acute ischemicstroke) may occur at a much earliertime—as early as 1–2 days in humansgiven intravenous recombinant tissueplasminogen activator less than 3 hoursafter stroke onset (40). Furthermore,Nagesh et al (41) demonstrated that al-though the mean ADC of an ischemiclesion is depressed within 10 hours, dif-ferent zones within an ischemic regionmay demonstrate low, pseudonormal, orelevated ADCs, suggesting different tem-poral rates of tissue evolution toward in-farction. Despite these variations, tissuecharacterized by an initial reduction inADC nearly always undergoes infarctionin humans.
DW and perfusion-weighted MR imagingfor assessment of stroke evolution.—Thecombination of perfusion-weighted andDW MR imaging may provide more in-formation than would either techniquealone. Perfusion-weighted imaging in-volves the detection of a decrease in sig-nal intensity as a result of the susceptibil-ity or T2* effects of gadolinium duringthe passage of a bolus of a gadolinium-based contrast agent through the intra-cranial vasculature (42,43). A variety ofhemodynamic images may be construc-ted from these data, including relativecerebral blood volume, relative cerebralblood flow, mean transit time, and time-to-peak maps (43–47).
In the context of arterial occlusion,brain regions with decreased diffusionand perfusion are thought to representnonviable tissue or the core of an infarc-
tion (31,32,34,39,48–51). The majorityof stroke lesions increase in volume onDW images, with the maximum volumeachieved at 2–3 days.
When most patients with acute strokeare evaluated with both DW and perfu-sion-weighted MR imaging, their imagesusually demonstrate one of three pat-terns (39,49–52): A lesion is smaller onDW images than the same lesion is onperfusion-weighted images; a lesion onDW images is equal to or larger than thaton perfusion-weighted images; or a le-sion is depicted on DW images but is notdemonstrable on perfusion-weighted im-ages. In large-vessel stroke lesions (suchas in the proximal portion of the middlecerebral artery), the abnormality as de-picted on perfusion-weighted images isfrequently larger than the lesion as de-picted on DW images. The peripheral re-gion, characterized by normal diffusionand decreased perfusion, usually pro-gresses to infarction unless there is earlyreperfusion. Thus, in the acute setting,perfusion-weighted imaging in combina-tion with DW imaging helps identify anoperational “ischemic penumbra” or areaat risk for infarction (Fig 6).
On the other hand, in small-vesselinfarctions (perforator infarctions anddistal middle cerebral artery infarc-tions), the initial lesion volumes onperfusion-weighted and DW imagesare usually similar, and the diffusion-weighted image lesion volume in-creases only slightly with time. A lesionlarger on DW images than on perfusion-weighted images or a lesion visible onDW images but not on perfusion-weighted images usually occurs withearly reperfusion. In this situation, thelesion on DW images usually does notchange substantially over time.
In animals treated with neuroprotec-tive agents after occlusion of the mid-dle cerebral artery, the increase instroke lesion volume on serial DW im-ages is reduced (53,54). This effect hasnot been convincingly demonstrated inhumans.
Reversibility of ischemic lesions on DWimages.—In animal models of ischemia,both a time threshold and an ADCthreshold for reversibility have beendemonstrated. In general, when the mid-dle cerebral artery in animals is tempo-rarily occluded for an hour or less, thediffusion lesion size markedly decreasesor resolves; however, when the middlecerebral artery is occluded for 2 hours ormore, the lesion size remains the same orincreases (17,34,55–57). Hasegawa et al(55) demonstrated that after 45 minutes
of temporary occlusion of the middle ce-rebral artery in rats, diffusion lesions arepartially or completely reversible whenthe difference in ADC values between theischemic region and a contralateral ho-mologous nonischemic region is notgreater than a threshold of 20.25 3 1025
cm2/sec. When the ADC difference isgreater than this threshold, the lesionnearly always becomes completely in-farcted. Similarly Dardzinski et al (58)demonstrated a threshold ADC of 0.55 31023 mm2/sec at 2 hours in a permanent-occlusion rat model.
In humans, reversibility of ischemic le-sions is rare. To our knowledge, only onecase has been reported in the literature(59), and we have observed reversibilityof only one ischemic lesion in over 2,000patients imaged in our clinical practice(Fig 7). That patient was treated with in-travenous recombinant tissue plasmino-gen activator 2 hours after symptomonset, and the initial ADC was approxi-mately 20% below that of contralateralhomologous nonischemic brain tissue. Inhumans, neither a threshold time nor athreshold ADC for reversibility have beenestablished.
DW imaging reliability in acute stroke.—Conventional computed tomography(CT) and MR imaging cannot be used toreliably detect infarction at the earliesttime points. The detection of hypoat-tenuation on CT scans and hyperinten-sity on T2-weighted MR images requires asubstantial increase in tissue water. Forinfarctions imaged within 6 hours afterstroke onset, reported (60,61) sensitivi-ties are 38%–45% for CT and 18%–46%for MR imaging. For infarctions imagedwithin 24 hours, the authors of one study(62) reported a sensitivity of 58% for CTand 82% for MR imaging.
DW images are very sensitive and spe-cific for the detection of hyperacute andacute infarctions, with a sensitivity of88%–100% and a specificity of 86%–100% (59,60,63). A lesion with decreaseddiffusion is strongly correlated with irre-versible infarction. Acute neurologic def-icits suggestive of stroke but without re-stricted diffusion are typically due totransient ischemic attack, peripheral ver-tigo, migraine, seizures, intracerebralhemorrhage, dementia, functional disor-ders, amyloid angiopathy, and metabolicdisorders (59,60,63).
Although, after 24 hours, infarctionsusually can be detected as hypoattenuat-ing lesions on CT and hyperintense le-sions on T2-weighted and fluid-attenu-ated inversion recovery MR images, DWimaging is useful in this setting, as well.
Volume 217 z Number 2 Diffusion-weighted MR Imaging of the Brain z 335
Older patients commonly have hyperin-tense abnormalities on T2-weighted im-ages that may be indistinguishable fromacute lesions. However, acute infarctionsare hyperintense on DW images andhypointense on ADC maps, whereaschronic foci are usually isointense onDW images and hyperintense on ADCmaps due to elevated diffusion (Fig 8). Inone study (64) in which there were indis-tinguishable acute and chronic whitematter lesions on T2-weighted images in69% of patients, the sensitivity and spec-ificity of DW imaging for detection of
acute subcortical infarction were 94.9%and 94.1%, respectively.
False-negative DW images have beenreported in patients with very small la-cunar brainstem or deep gray nuclei in-farction (60,63,65). Some of these le-sions were seen on follow-up DWimages, and others were presumed to bepresent on the basis of clinical deficits.False-negative DW images also occur inpatients with regions of decreased per-
fusion (increased mean transit time anddecreased relative cerebral blood flow),which are hyperintense on follow-upDW images; in other words, these pa-tients initially had regions character-ized by ischemic but viable tissue thatprogressed to infarction. These findingsstress the importance of obtaining earlyfollow-up images in patients with nor-mal DW images and persistent stroke-like deficits, so that infarctions or areas
Figure 5. Time course of an ischemic infarction. Images demon-strate the evolution of an ischemic infarction involving the leftcerebellar hemisphere and left middle cerebellar peduncle. Bothtransverse DW images (DWI; b 5 1,000 sec/mm2; effective gradient,14 mT/m; repetition time msec/echo time msec, 6,000/108; matrix,256 3 128; field of view, 400 3 200 mm; section thickness, 6 mmwith 1-mm gap) and transverse ADC maps are displayed. The patientunderwent MR imaging 6 hours after the onset of acute neurologicsymptoms. At 6 hours, the lesion (arrows) is hyperintense on the DWimages and hypointense on the corresponding ADC map. The lesionbecomes progressively more hyperintense on DW images, reachingits maximum hyperintensity at the 58-hour time point, when it alsoreaches its maximum hypointensity on ADC maps. At 7 days, there isongoing resolution of the lesion on both DW images and ADC maps.By 134 days, there is subtle hypointensity on the DW image andhyperintensity on the ADC images.
Figure 6. Diffusion-perfusion mismatch after left middle cerebralartery stroke. The patient was imaged 3.8 hours after a witnessedsudden onset of a right hemiparesis. Transverse DW images (DWI; b 51,000 sec/mm2; effective gradient, 14 mT/m; 6,000/108; matrix,256 3 128; field of view, 400 3 200 mm; section thickness, 6 mmwith 1-mm gap) demonstrate hyperintensity in the subcortical re-gion, including in the lenticular nucleus and corona radiata (arrow-heads, right-hand image in top row). Transverse cerebral blood vol-ume (CBV) images (spin-echo echo-planar technique; 0.2 mmol/kggadopentetate dimeglumine [Magnevist; Berlex Laboratories, Wayne,NJ]; 51 images per section; 1,500/75; matrix, 256 3 128; field of view,400 3 200 mm; section thickness, 6 mm with 1-mm gap) demon-strate decreased dynamic cerebral blood volume in the region ofhyperintensity on the DW images. However, there are areas of abnor-mal cerebral blood volume (arrows) that appear relatively normal onthe DW study. Follow-up study performed 10 hours after the onset ofsymptoms demonstrates an increase in the size of the DW imagingabnormality (arrowheads, right-hand image in fifth row) as it extendsinto the region of brain that was previously normal on DW imagesbut abnormal on cerebral blood volume images.
336 z Radiology z November 2000 Schaefer et al
at risk for infarction are identified andtreated as early as possible.
False-positive DW images have beenreported in patients with a diagnosisother than acute infarction. These in-clude cerebral abscess (with restricteddiffusion on the basis of viscosity) andtumor (with restricted diffusion on thebasis of dense cell packing). Whenthese lesions are viewed on DW imagesin combination with other routine T1-and T2-weighted MR images, they canusually be differentiated from acute in-farctions.
Correlation of DW MR imaging with clin-ical outcome.—DW MR imaging findingsmay reflect the severity of clinical neuro-logic deficits and help predict clinicaloutcome. Statistically significant correla-tions between the acute DW MR lesionvolume and both acute and chronicneurologic assessment results, including
those of the National Institutes of HealthStroke Score Scale, the Canadian Neuro-logic Scale, the Barthel Index, and theRankin Scale, have been demonstrated(39,51,66–68). This correlation is stron-ger in cases of cortical stroke and weakerin cases of penetrator artery stroke (39,66).Lesion location likely explains the vari-ance; for example, a lesion in a majorwhite matter tract may produce a moreprofound neurologic deficit than would acortical lesion of the same size. There alsois a weaker correlation between initial le-sion volume and National Institutes ofHealth Stroke Score Scale measures in pa-tients with a prior infarction. In addition,there is a significant correlation betweenthe acute ADC ratio (lesion ADC to nor-mal contralateral brain ADC) and chronicneurologic assessment scale scores (39,68).Perfusion-weighted image volumes alsocorrelate with acute and chronic neuro-
logic assessment test results (51,67). In onestudy (51), patients who had lesion vol-umes on perfusion-weighted images thatwere larger than volumes on DW images(perfusion-diffusion mismatches) hadworse outcomes and larger final infarctvolumes. In another study (39), patientswith early reperfusion had smaller finalinfarct volumes and better clinical out-comes. Because DW and perfusion-weighted MR imaging can help predictclinical outcome at very early time points,these techniques may prove to be valuablefor the selection of patients for thrombol-ysis or administration of neuroprotectiveagents.
Neonatal hypoxic ischemic brain inju-ry.—DW MR imaging is rapidly improv-ing the evaluation of neonatal hypoxicischemic encephalopathy and focal in-farctions. Animal models of neonatalischemia have demonstrated lesions onDW MR images as early as 1 hour afterligation of the carotid artery (69,70). Inhumans, within 1 day of birth, acute is-chemic lesions not seen on routine CT orMR images are identified on DW MR im-ages (71,72). When lesions are identifiedon conventional images, lesion conspi-cuity is increased and lesion extent isseen to be larger on DW MR images. Inaddition, lesions identified on the initialDW MR images are identified on fol-low-up conventional images and, there-fore, help accurately predict the extent ofinfarction. This correlates with the find-ing in animals that areas of restricted dif-fusion correlate with areas of injury atautopsy.
Animal models have also demon-strated the evolution of neonatal hy-poxic ischemic injury over time. In a rab-bit model (70), ischemic lesions wereseen first in the cortex, followed by thesubcortical white matter, the ipsilateralbasal ganglia, and the contralateral basalganglia.
Thus, DW MR imaging is helping in-crease our understanding of the patho-physiology of neonatal ischemia. It al-lows timing of ischemic onset, providesearlier and more reliable detection ofacute ischemic lesions, and allows differ-entiation of focal infarctions from moreglobal hypoxic ischemic lesions. This in-formation may provide a better early as-sessment of the long-term prognosis andmay be important in the evaluation ofnew neuroprotective agents.
Transient ischemic attacks.—Nearly 50%of patients with transient ischemic at-tacks have lesions characterized by re-stricted diffusion (73,74). These lesionsare usually small (,15-mm diameter), are
Figure 7. Reversible ischemic lesion. Top: The patient was imaged approximately 2 hours afterthe onset of a witnessed acute neurologic deficit. Top left: Transverse DW image (DWI; b 5 1,000sec/mm2; effective gradient, 14 mT/m; 6,000/108; matrix, 256 3 128; field of view, 400 3 200mm; section thickness, 6 mm with 1-mm gap) shows an area of hyperintensity (arrow) in the leftposterior frontal and anterior parietal lobes. Top middle: A region of hypointensity (arrow)corresponding to this area is seen on the transverse ADC image (arrow). Top right: No definiteabnormality is seen on the transverse fast spin-echo T2-weighted MR image (4,000/104; echotrain length, eight; matrix, 256 3 192; field of view, 200 3 200 mm; section thickness, 5 mm with1-mm gap; one signal acquired). The patient was treated with intravenous recombinant tissueplasminogen activator, with resolution of the neurologic symptoms. Bottom: Follow-up imagesobtained 3 days later demonstrate near interval resolution of the abnormalities on the 2-hour DWimage and ADC map. No definite lesion was identified on the follow-up T2-weighted image. Ofnote, the decrease in ADC was approximately 20% of the normal value. Lesions that becomeconfirmed infarctions typically demonstrate a 50% reduction in ADC.
Volume 217 z Number 2 Diffusion-weighted MR Imaging of the Brain z 337
almost always in the clinically expectedvascular territory, and are thought to rep-resent markers of more widespread re-versible ischemia. In one study (74), 20%of the lesions were not seen at follow-up;the lesions could have been reversible or,owing to atrophy, too small to see onconventional MR images. The informa-tion obtained from DW MR imagingchanged the suspected localization of anischemic lesion, as well as the suspectedetiologic mechanism, in more than one-third of patients (74). In another study(73), statistically significant independentpredictors for identification of these le-sions on DW MR images included previ-ous nonstereotypic transient ischemic at-tack, cortical syndrome, or an identifiedstroke mechanism, and the authors sug-gested an increased stroke risk in patientswith these lesions. Early identification ofpatients with transient ischemic attackwith increased risk of stroke and betteridentification of etiologic mechanisms ischanging acute management and may af-fect patient outcome.
Other clinical stroke mimics.—These syn-dromes generally fall into two categories:(a) nonischemic lesions with no acuteabnormality on routine or DW MR im-ages or (b) vasogenic edema syndromesthat mimic acute infarction on conven-tional MR images. Nonischemic syn-dromes with no acute abnormality iden-tified on DW or conventional MR imagesand reversible clinical deficits includeperipheral vertigo, migraines, seizures,dementia, functional disorders, amy-loid angiopathy, and metabolic disor-ders (59,60,63). When a patients withthese syndromes present, we can confi-dently predict that they are not under-going infarction; they are spared unnec-essary anticoagulation treatment and astroke work-up.
Syndromes with potentially revers-ible vasogenic edema include eclampsia, hy-pertensive encephalopathy, cyclosporintoxicity, other posterior leukoencepha-lopathies, venous thrombosis, humanimmunodeficiency virus encephalopa-thy, and hyperperfusion syndrome aftercarotid endarterectomy (Fig 9). Patientswith these syndromes frequently presentwith neurologic deficits that are sugges-tive of acute ischemic stroke or with neu-rologic deficits such as headache orseizure that are suggestive of vasogenicedema, but ischemic stroke is still astrong diagnostic consideration. Con-ventional MR imaging cannot help dif-ferentiate vasogenic edema from the cy-totoxic edema associated with acuteinfarction. Cytotoxic edema produces
high signal intensity in gray and/or whitematter on T2-weighted images. Althoughvasogenic edema on T2-weighted imagestypically produces high signal intensityin white matter, the hyperintensity can in-volve adjacent gray matter. Consequently,posterior leukoencephalopathy can some-times mimic infarction of the posteriorcerebral artery. Hyperperfusion syndromeafter carotid endarterectomy can resembleinfarction of the middle cerebral artery.Human immunodeficiency virus encepha-lopathy can produce lesions in a variety ofdistributions, some of which have a mani-festation similar to that of arterial infarc-tion. Deep venous thrombosis can producebilateral thalamic hyperintensity that is in-distinguishable from “top of the basilar”syndrome arterial infarction.
DW MR imaging can be used to reli-ably distinguish vasogenic from cyto-toxic edema. Whereas cytotoxic edema ischaracterized by restricted diffusion, va-sogenic edema is characterized by ele-vated diffusion due to a relative increasein water in the extracellular compart-ment, where water is more mobile (75–78). On DW MR images, vasogenic
edema may be hypointense to slightlyhyperintense, because these images haveboth T2 and diffusion contributions.When vasogenic edema is hyperintenseon DW MR images, it can mimic hyper-acute or subacute infarction. On ADC im-ages, cytotoxic edema due to ischemia isalways hypointense for 1–2 weeks, andvasogenic edema is always hyperintense.Therefore, DW MR images should alwaysbe compared with ADC images.
Correct differentiation of vasogenic fromcytotoxic edema affects patient care. Mis-diagnosis of vasogenic edema syndromeas acute ischemia could lead to unneces-sary and potentially dangerous use ofthrombolytics, antiplatelet agents, antico-agulants, and vasoactive agents. Further-more, failure to correct relative hyperten-sion could result in increased cerebraledema, hemorrhage, seizures, or death.Misinterpretation of acute ischemic in-farction as vasogenic edema syndromewould discourage proper treatment withanticoagulants, evaluation for an em-bolic source, and liberal blood pressurecontrol, which could increase the risk ofrecurrent brain infarction.
Figure 8. Differentiation of acute white matter infarction from nonspecific small-vesselischemic changes. This patient had onset of symptoms 2 days prior to imaging. Top: Trans-verse DW images (DWI; b 5 1,000 sec/mm2; effective gradient, 14 mT/m; 6,000/108; matrix,256 3 128; field of view, 400 3 200 mm; section thickness, 6 mm with 1-mm gap) in the toprow clearly demonstrate the acute infarction (arrowheads) in the putamen and coronaradiata. Bottom: Fluid-attenuated inversion recovery (FLAIR) images (10,000/141; inversiontime, 2,200 msec; echo train length, eight; matrix, 256 3 192; field of view, 240 3 240 mm;section thickness, 5 mm with 1-mm gap; one signal acquired) demonstrate multiple whitematter lesions in which acute (arrowhead) and chronic lesions (arrows) cannot be differen-tiated.
338 z Radiology z November 2000 Schaefer et al
Masses
Extraaxial masses: arachnoid cyst versusepidermoid tumor.—Conventional MR im-ages cannot be used to reliably distin-guish epidermoid tumors from arachnoid
cysts; both lesions are very hypointenserelative to brain parenchyma on T1-weighted MR images and very hyperin-tense on T2-weighted images. Epider-moid tumors are solid masses, however,which demonstrate ADCs similar to
those of gray matter and lower thanthose of CSF (79,80). With the combina-tion of T2 and diffusion effects, epider-moid tumors are markedly hyperintensecompared with CSF and brain tissue ondiffusion MR images. Conversely, arach-noid cysts are fluid filled, demonstratevery high ADCs, and appear similar toCSF on DW MR images. Furthermore, onconventional MR images obtained afterresection of an epidermoid tumor, theresection cavity and residual tumor maybe similarly hypointense on T1-weightedimages and hyperintense on T2-weightedimages. On DW MR images, the hypoin-tense CSF-containing cavity can easily bedifferentiated from the residual hyperin-tense epidermoid tumor (Fig 10).
Intraaxial masses.—A number of investi-gators (76,81–86) have evaluated DW MRimaging of intraaxial tumors (predomi-nantly gliomas) in animals and humans. Ithas been demonstrated (76,81,83) that tu-mor and edema have higher ADCs thandoes normal brain tissue and that centralnecrosis has a higher ADC than do tumor,edema, or normal brain tissue. Tien et al(76) demonstrated that enhancing tumorshave significantly lower ADCs than doesadjacent edema, but Brunberg et al (81)found that there is no significant differ-ence between ADCs of enhancing tumorand edema. Both concluded that the ADCalone cannot be used to differentiate anonenhancing tumor from adjacentedema. Brunberg et al suggested that bothenhancing and nonenhancing tumors canbe distinguished from edema becauseedema has significantly higher indices ofdiffusion anisotropy when compared withadjacent tumor, presumably due to intactmyelin within the edema. Demarcationof tumor from surrounding vasogenicedema with DW MR imaging may be im-portant in determining radiation ports,surgical margins, and biopsy sites. A num-ber of investigators (81,84,85) have dem-onstrated that DW MR imaging cannotbe used to differentiate between high-and low-grade gliomas or between tumortypes.
DW MR imaging is also valuable inthe assessment of tumor resections thatare complicated, in the immediate post-operative period, by acute neurologicdeficits. Although both extracellular edemaand infarction are hyperintense on spin-echo T2-weighted images, cytotoxicedema is characterized by a low ADC,and vasogenic edema is characterized bya high ADC, relative to brain paren-chyma. Thus, an acute infarction can eas-ily be differentiated from postoperativeedema.
Figure 9. Hyperperfusion syndrome after carotid endarterectomy. Thepatient developed neurologic symptoms referable to the left hemisphereseveral days after undergoing a left carotid endarterectomy. The CT scanwas abnormal, and the question of infarction was raised. Left: Transversefast spin-echo T2-weighted MR image (4,000/104; echo train length,eight; matrix, 256 3 192; field of view, 200 3 200 mm, section thickness,5 mm with 1-mm gap; one signal acquired) demonstrates numerousareas of abnormal high signal intensity (arrow) in the left hemisphere.Infarctions remained in the differential diagnosis. Middle: TransverseDW MR image (b 5 1,000 sec/mm2; effective gradient, 14 mT/m; 6,000/108; matrix, 256 3 128; field of view, 400 3 200 mm; section thickness,6 mm with 1-mm gap) reveals predominant isointensity in the lefthemisphere with small areas of slight hypointensity and slight hyperin-tensity (arrow). ADC images (not shown) demonstrated no areas ofrestricted diffusion. Right: Transverse three-dimensional time-of-flightMR angiogram (49/6.9; 20° flip angle; matrix, 256 3 192; field of view,200 3 200 mm; section thickness, 1 mm) demonstrates excellent flow-related enhancement (arrow) in the left hemisphere. A diagnosis ofhyperperfusion syndrome with vasogenic edema was established on thebasis of DW imaging findings. The patient was treated conservativelyand recovered fully.
Figure 10. Postoperative residual epidermoid tumor. The patientunderwent resection of a large left middle cranial fossa epidermoidtumor that extended into the posterior fossa. Transverse T1-weighted(left) (650/16; matrix, 256 3 192; field of view, 200 3 200 mm;section thickness, 5 mm with 1-mm gap; one signal acquired) and fastspin echo T2-weighted (middle) (4,000/104; echo train length, eight;matrix, 256 3 192; field of view, 200 3 200 mm; section thickness, 5mm with 1-mm gap, one signal acquired) MR images do not allowclear differentiation of residual mass from the resection cavity. Right:Transverse DW MR image (b 5 1,000 sec/mm2; effective gradient, 14mT/m; 6,000/108; matrix, 256 3 128; field of view, 400 3 200 mm;section thickness, 6 mm with 1-mm gap) clearly demonstrates ahyperintense mass (black arrow) adjacent to the left pons and asmaller amount of mass (white arrow) in the left middle cranial fossa,consistent with residual epidermoid tumor. CSF (arrowhead) in theresection cavity is markedly hypointense.
Volume 217 z Number 2 Diffusion-weighted MR Imaging of the Brain z 339
Intracranial Infections
Pyogenic infection.—Abscess cavities andempyemas are homogeneously hyperin-tense on DW MR images (Fig 11), withsignal intensity ratios of abscess cavity tonormal brain tissue that range from 2.5 to6.9 and with ADC ratios that range from0.36 to 0.46 (87–89). In one study (88), theADC of the abscess cavity in vivo was sim-ilar to that of pus aspirated from the cavityin vitro. In another study (89), the ADCratio of empyema compared with CSF was0.13 in one patient. The relatively re-stricted diffusion most likely results fromthe high viscosity and cellularity of pus.
Although intracranial abscesses andintracranial neoplasms may appear sim-ilar on images obtained with conven-tional MR sequences, the signal inten-sity of the abscess cavity is markedlyhigher and the ADC ratios are lowerthan those of necrotic tumors on DWMR images (76,81,87). Bacterial menin-gitis may be complicated by subduraleffusions or subdural empyemas, whichare difficult to differentiate on conven-tional MR images. Empyemas are hy-perintense on DW MR images and havea restricted ADC, whereas sterile effu-sions are hypointense and have an ADCsimilar to that of CSF. Thus, DW MRimages may be important when decid-ing whether to drain or conservativelymanage extraaxial collections associ-ated with meningitis.
Herpes encephalitis.—Herpes encepha-litis lesions are characterized by markedhyperintensity on DW MR images (Fig12), with ADC ratios of these lesionsto normal brain parenchyma rangingfrom 0.48 to 0.66. On follow-up conven-tional T1-weighted and T2-weighted MRimages, these areas demonstrate ence-phalomalacic change. The restricted dif-fusion is explained by cytotoxic edema intissue undergoing necrosis. DW MR imag-ing may aid in distinguishing herpeslesions from infiltrative temporal lobetumors because the ADCs of herpes le-sions are low while the ADCs of varioustumors are elevated or in the normalrange (76,81).
Creutzfeldt-Jakob disease.—DW MR im-ages in patients with Creutzfeldt-Jakobdisease have demonstrated hyperintenselesions in the cortex and basal ganglia(Fig 13). ADCs in lesions in five patientswere significantly lower than those ofnormal brain parenchyma (90,91), whileADCs in lesions in two patients werenormal or mildly elevated (92). Thevariable ADCs are likely related to vari-
able amounts of spongiform change,neuronal loss, and gliosis.
Whereas Creutzfeldt-Jakob disease isclassically characterized by progressivedementia, myoclonic jerks, and periodicsharp-wave electroencephalographic ac-tivity, these features frequently are ab-sent, and Creutzfeldt-Jakob disease can-not be clinically distinguished fromother dementing illnesses (93,94). Fur-thermore, conventional MR images maybe normal in as many as 21% of patients(95). Thus, DW MR imaging may beuseful for help in the diagnosis ofCreutzfeldt-Jakob disease and in the dif-ferentiation from Alzheimer disease.
Trauma
Results of an experimental study (96)of head trauma have demonstrated thatmoderate fluid-percussion injury leads toincreased diffusion, reflecting increasedextracellular water, in rat cortex and hip-pocampus. This correlates with a report(97) that moderate fluid-percussion in-jury does not reduce cerebral blood flowenough to induce ischemia. Ito et al (98)demonstrated no significant change inbrain ADCs when rats are subjected toimpact acceleration trauma alone. How-ever, when trauma is coupled with hyp-oxia and hypotension, the ADCs in ratcortex and thalami decrease significantly
Figure 11. Pathologically proved cerebral abscess. Left: A complexsignal intensity pattern is visible in the right occipital and temporallobes on the fast spin-echo T2-weighted MR image (4,000/104; echotrain length, eight; matrix, 256 3 192; field of view, 200 3 200 mm;section thickness, 5 mm with 1-mm gap; one signal acquired). Mid-dle: Ring-enhancing lesion (arrows) in the right occipital lobe isdemonstrated on the gadolinium-enhanced T1-weighted MR image(650/16; matrix, 256 3 192; field of view, 200 3 200 mm; sectionthickness, 5 mm with 1-mm gap; one signal acquired). Right: DW MRimage (b 5 1,000 sec/mm2; effective gradient, 14 mT/m; 6,000/108;matrix, 256 3 128; field of view, 400 3 200 mm; section thickness, 6mm with 1-mm gap) demonstrates the characteristic restricted diffu-sion of pyogenic abscess (arrows). Note the hyperintensity (arrow-head) in the left occipital horn due to a loculated collection of pus inthis location.
Figure 12. Herpes encephalitis proved with results of polymerasechain reaction test. DW MR images (DWI; b 5 1,000 sec/mm2; effec-tive gradient, 14 mT/m; 6,000/108; matrix, 256 3 128; field of view,400 3 200 mm; section thickness, 6 mm with 1-mm gap). demon-strate restricted diffusion bilaterally in the temporal lobes (shortarrows), inferior frontal lobes (long arrows), and insulae (arrow-heads), which is a typical distribution for herpes encephalitis.
340 z Radiology z November 2000 Schaefer et al
and neuronal injury was observed histo-logically. They concluded that brain isch-emia associated with severe head traumaleads to cytotoxic edema. Barzo et al (99)demonstrated a reduction in rat brainADCs hours to weeks after an impact ac-celeration injury. They concluded thatcerebral blood flow does not decreaseenough to cause ischemic edema andthat neurotoxic edema causes the re-duced ADCs and neuronal injury.
DW MR imaging in 116 diffuse axonalinjury lesions in humans (100) demon-strated changes similar to those in ani-mal models: ADCs were reduced in 64%of lesions, were elevated in 34%, andwere similar to ADCs of normal braintissue in 12%. In addition, most lesionswere more conspicuous on DW MR im-ages than on conventional T2-weightedimages (Fig 14). Thus, DW MR imagingmay be important for the prospective de-termination of the extent of traumaticinjury, the degree of irreversible injury(number of lesions characterized by lowADCs indicative of cytotoxic edema), andthe long-term prognosis.
Hemorrhage
The appearance of hemorrhage on DWMR images is complex and involves
many factors, including the relativeamounts of different hemorrhagic prod-ucts and the pulse sequence used (Fig 15).Oxyhemoglobin is hyperintense on DWimages and has a lower ADC than doesnormal brain tissue; this may indicatethe relative restriction of water move-ment inside the red blood cell (101). Ex-tracellular methemoglobin has a higherADC than does normal brain tissue,which indicates that the mobility of wa-ter in the extracellular space is increased.The prolongation of the T2 componentof fluid with extracellular methemoglo-bin results in hyperintensity on DW im-ages. Hemorrhage containing deoxyhe-moglobin, intracellular methemoglobin,and hemosiderin are hypointense on DWimages because of magnetic susceptibilityeffects. Because these products of hemor-rhage have very low signal intensity onT2-weighted images, ADCs cannot be re-liably calculated for them.
Demyelination
Multiple sclerosis.—In animals with ex-perimental allergic encephalomyelitis (amodel of multiple sclerosis) and in pa-tients with multiple sclerosis, most plaquesdemonstrate increased diffusion (102–106).In humans, acute plaques have signifi-
cantly higher ADCs than do chronicplaques (105,106). The elevated diffusionmay result from an increase in the size ofthe extracellular space due to edema anddemyelination acutely and to axonal lossand gliosis chronically. In rare instances,acute plaques have restricted diffusion.This may result from increased inflam-matory cellular infiltration with littleextracellular edema. Of interest, normal-appearing white matter has a mildly in-creased ADC (104). This correlates withhistologic results in which multiple scle-rosis was shown to diffusely affect whitematter (107).
In monkeys with experimental allergicencephalomyelitis, Heide et al (102)demonstrated that diffusion anisotropydecreased over time. We have also ob-served this phenomenon in humans. Fur-thermore, Verhoye et al (103) demon-strated a significant positive correlationbetween the degree of ADC elevation inthe external capsule and severity of clin-ical disease in rats with experimental al-lergic encephalomyelitis. However, thisrelationship has not been confirmed inhumans. Horsfield et al (104) demon-strated that benign multiple sclerosis le-sions have ADCs similar to those of sec-ondary progressive multiple sclerosis.Furthermore, the degree of ADC eleva-tion within individual lesions did notcorrelate with the degree of patient dis-ability.
Acute disseminated encephalomyelitis.—Acute disseminated encephalomyelitis le-sions have ADCs higher than those of nor-mal white matter, likely as a result ofdemyelination and increased extracellularwater. DW MR imaging cannot help distin-guish between multiple sclerosis and acutedisseminated encephalomyelitis lesionsbecause both usually have elevated diffu-sion. Because acute infarctions are charac-terized by restricted diffusion, however,DW MR imaging should be reliable forhelp in the differentiation between demy-elinating lesions and stroke.
CONCLUSION
The DW MR pulse sequence is a valuabletechnique. It provides information onthe physiologic state of the brain and isparticularly sensitive to ischemic infarc-tion. We recommend its use when thereis an acute neurologic deficit. As DW im-aging improves and becomes more wide-spread, it is expected to play a greater rolein the diagnosis of hyperacute and acutestroke and in the differentiation of strokefrom other disease processes that mani-
Figure 13. Pathologically proved Creutzfeldt-Jakob disease. Top: Transverse T2-weighted MRimages (4,000/104; echo train length, eight; matrix, 256 3 192; field of view, 200 3 200 mm;section thickness, 5 mm with 1-mm gap; one signal acquired) demonstrate hyperintensity of thebasal ganglia. Bottom: Transverse DW MR images (DWI; b 5 1,000 sec/mm2; effective gradient,14 mT/m; 6,000/108; matrix, 256 3 128; field of view, 400 3 200 mm; section thickness, 6 mmwith 1-mm gap) show marked hyperintensity involving the basal ganglia bilaterally (arrowheads)and portions of the bilateral cortical ribbon (arrows).
Volume 217 z Number 2 Diffusion-weighted MR Imaging of the Brain z 341
fest with acute neurologic deficits. DWMR imaging will also play a greater rolein the management of stroke and maybe helpful in the selection of patientsfor thrombolysis and in the evaluationof new neuroprotective agents. It mayprove to be valuable in the evaluation ofa wide variety of other disease processes,as described in this review.
Acknowledgments: The authors thank Nan-dita Guha-Thakurta, MD, and Cara O’Reilly,BA, for their help in the preparation of thismanuscript.
References1. Moseley ME, Wendland MF, Kucharczyk
J. Magnetic resonance imaging of diffu-sion and perfusion. Top Magn Reson Im-aging 1991; 3:50–67.
2. Warach S, Chien D, Li W, Ronthal M,Edelman RR. Fast magnetic resonancediffusion-weighted imaging of acute hu-man stroke. Neurology 1992; 42:1717–1723. [Erratum: Neurology 1992; 42:2192]
3. Hossmann KA, Fischer M, Bockhorst K,Hoehn-Berlage M. NMR imaging of theapparent diffusion coefficient (ADC) forthe evaluation of metabolic suppressionand recovery after prolonged cerebralischemia. J Cereb Blood Flow Metab1994; 14:723–731.
4. Le Bihan D. Diffusion and perfusionmagnetic resonance imaging: applica-tions to functional MRI. New York, NY:Raven, 1995.
5. Conturo TE, McKinstry RC, AronovitzJA, Neil JJ. Diffusion MRI: precision, ac-curacy and flow effects. NMR Biomed1995; 8:307–332.
6. Stejskal E, Tanner J. Spin diffusion mea-surements: spin echos in the presence of
time-dependent field gradient. J ChemPhys 1965; 42:288–292.
7. Liu G, van Gelderen P, Duyn J, MoonenCT. Single-shot diffusion MRI of humanbrain on a conventional clinical instru-ment. Magn Reson Med 1996; 35:671–677.
9. Gudbjartsson H, Maier SE, Mulkern RV,Morocz IA, Patz S, Jolesz FA. Line scandiffusion imaging. Magn Reson Med1996; 36:509–519.
10. Li TQ, Takahashi AM, Hindmarsh T,Moseley ME. ADC mapping by means ofa single-shot spiral MRI technique withapplication in acute cerebral ischemia.Magn Reson Med 1999; 41:143–147.
11. Brockstedt S, Thomsen C, Wirestam R,Holtas S, Stahlberg F. Quantitative diffu-sion coefficient maps using fast spin-echo MRI. Magn Reson Imaging 1998;16:877–886.
12. de Crespigny AJ, Marks MP, EnzmannDR, Moseley ME. Navigated diffusionimaging of normal and ischemic humanbrain. Magn Reson Med 1995; 33:720–728.
13. Nomura Y, Sakuma H, Takeda K, TagamiT, Okuda Y, Nakagawa T. Diffusional an-isotropy of the human brain assessedwith diffusion-weighted MR: relationwith normal brain development and ag-ing. AJNR Am J Neuroradiol 1994; 15:231–238.
Figure 14. Severe head trauma resulting in diffuse axonal injury.Top: Transverse T2-weighted MR images (4,000/104; echo trainlength, eight; matrix, 256 3 192; field of view, 200 3 200 mm; sectionthickness, 5 mm with 1-mm gap; one signal acquired) demonstratemultiple white matter hyperintensities (arrows). Bottom: TransverseDW MR images (DWI; b 5 1,000 sec/mm2; effective gradient, 14mT/m; 6,000/108; matrix, 256 3 128; field of view, 400 3 200 mm;section thickness, 6 mm with 1-mm gap) demonstrate the lesions(arrows) with increased conspicuity. The hyperintensity is consistentwith restricted diffusion. Note abnormalities (arrowheads) that ex-tend to the cortex posteriorly.
Figure 15. Hematoma in a patient with a right hemisphere glioblas-toma who had undergone prior resection and who had developed ahematoma in the right frontal lobe. The patient was hospitalized forprogression of symptoms and development of fever. A ring-enhanc-ing lesion at the site of the prior hematoma was seen on a gadolini-um-enhanced T1-weighted MR image (not shown) in the right frontallobe. Left: DW MR image (b 5 1,000 sec/mm2; effective gradient, 14mT/m; 6,000/108; matrix, 256 3 128; field of view, 400 3 200 mm;section thickness, 6 mm with 1-mm gap) demonstrates a hyperin-tense lesion (arrow) in the right frontal lobe. Right: On the ADCimage, the lesion is hypointense (arrow), which is consistent withrestricted diffusion. The lesion was drained, and old hemorrhage wasdemonstrated. There was no evidence of infection.
342 z Radiology z November 2000 Schaefer et al
fusion imaging of cerebral infarction inhumans. AJNR Am J Neuroradiol 1992;13:1097–1102.
16. Mintorovitch J, Yang GY, Shimizu H,Kucharczyk J, Chan PH, Weinstein PR.Diffusion-weighted magnetic resonanceimaging of acute focal cerebral ischemia:comparison of signal intensity withchanges in brain water and Na1,K(1)-ATPase activity. J Cereb Blood FlowMetab 1994; 14:332–336.
17. Kucharczyk J, Vexler ZS, Roberts TP, etal. Echo-planar perfusion-sensitive MRimaging of acute cerebral ischemia. Ra-diology 1993; 188:711–717.
18. Matsumoto K, Lo EH, Pierce AR, GarridoL, Kowall NW. Role of vasogenic edemaand tissue cavitation in ischemic evolu-tion on diffusion-weighted imaging:comparison with multiparameter MRand immunohistochemistry. AJNR Am JNeuroradiol 1995; 16:1107–1115.
19. Sevick R, Kanda F, Mintorovitch J, et al.Cytotoxic brain edema: assessment withdiffusion-weighted MR imaging. Radiol-ogy 1992; 185:687–690.
20. Benveniste H, Hedlund LW, JohnsonGA. Mechanism of detection of acutecerebral ischemia in rats by diffusion-weighted magnetic resonance micros-copy. Stroke 1992; 23:746–754.
21. Schuier G, Tigges A, Reimer P, DaldrupH, Peters P. The repeatability of MR per-fusion studies: a clinical study (abstr).In: Proceedings of the Third Meeting ofthe International Society for MagneticResonance in Medicine. Berkeley, Cali-fornia: International Society for Mag-netic Resonance in Medicine, 1995; 84.
22. Niendorf T, Dijkhuizen R, Norris D,Campagne M, Nicolay K. Biexponentialdiffusion attenuation in various states ofbrain tissue: implications for diffusionweighted imaging. Magn Reson Med1996: 847–857.
23. Sykova E, Svoboda J, Polak J, Chvatal A.Extracellular volume fraction and diffu-sion characteristics during progressiveischemia and terminal anoxia in the spi-nal cord of the rat. J Cereb Blood FlowMetab 1994; 14:301–311.
24. Wick M, Nagatomo Y, Prielmeier F,Frahm J. Alteration of intracellular me-tabolite diffusion in rat brain in vivoduring ischemia and reperfusion. Stroke1995; 26:1930–1934.
25. van der Toorn A, Dijkhuizen R, TullekenC, Nicolay K. Diffusion of metabolites innormal and ischemic rat brain measuredby localized 1H MRS. Magn Reson Med1996; 36:914–922.
26. Duong TQ, Ackerman JJ, Ying HS, NeilJJ. Evaluation of extra- and intracellularapparent diffusion in normal and glo-bally ischemic rat brain via 19F NMR.Magn Reson Med 1998; 40:1–13.
27. Morikawa E, Ginsberg M, Dietrich W, etal. The significance of brain temperaturein focal cerebral ischemia: histopatho-logical consequences of middle cerebralartery occlusion in the rat. J Cereb BloodFlow Metab 1992; 12:380–389.
28. Le Bihan D, Delannoy J, Levin RL.Temperature mapping with MR imag-ing of molecular diffusion: applicationto hyperthermia. Radiology 1989; 171:853–857.
29. Helpern J, Ordidige R, Knight R. The ef-
fect of cell membrane water permeabil-ity on the apparent diffusion coefficientof water (abstr). In: Book of abstracts:Society of Magnetic Resonance in Med-icine 1992. Berkeley, Calif: Society ofMagnetic Resonance in Medicine, 1992;12.
30. Szafer A, Zhong J, Gore J. Theoreticalmodel for water diffusion in tissues.Magn Reson Med 1995; 33:697–712.
31. Kucharczyk J, Mintorovitch J, Asgari H,Moseley M. Diffusion/perfusion MR im-aging of acute cerebral ischemia. MagnReson Med 1991; 19:311–315.
32. Moseley M, Kucharczyk J, MintorovitchJ, et al. Diffusion-weighted MR imagingof acute stroke: correlation with T2-weighted and magnetic susceptibility-enhanced MR imaging in cats. AJNRAm J Neuroradiol 1990; 11:423–429.
33. Moseley M. Early detection of regionalcerebral ischemia in cats: comparison ofdiffusion- and T2-weighted MRI andspectroscopy. Magn Reson Med 1990;14:330–346.
34. Mintorovitch J, Moseley M, Chileuitt L,Shimizu H, Cohen Y, Weinstein P. Com-parison of diffusion and T2-weightedMRI for the early detection of cerebralischemia and reperfusion in rats. MagnReson Med 1991; 18:39–50.
35. Moonen C, Pekar J, Vleeschouwer MD,Van Gelderen P, Zijl PV, DesPres D. Re-stricted and anisotropic displacement ofwater in healthy cat brain and in strokestudied by NMR diffusion imaging. MagReson Med 1991; 19:327–332.
36. Warach S, Gaa J, Siewert B, WielopolskiP, Edelman R. Acute human stroke stud-ied by whole brain echo planar diffu-sion-weighted magnetic resonance im-aging. Ann Neurol 1995; 37:231–241.
37. Schlaug G, Siewert B, Benfield A, Edel-man R, Warach S. Time course of theapparent diffusion coefficient (ADC) ab-normality in human stroke. Neurology1997; 49:113–119.
38. Lutsep H, Albers G, DeCrespigny A, KamatG, Marks M, Moseley M. Clinical utility ofdiffusion-weighted magnetic resonanceimaging in the assessment of ischemicstroke. Neurology 1997; 41:574–580.
39. Schwamm L, Koroshetz W, Sorensen A,et al. Time course of lesion developmentin patients with acute stroke: serial dif-fusion and hemodynamic weightedmagnetic resonance imaging. Stroke1998; 29:2268–2276.
40. Marks M, Tong DC, Beaulieu C, AlbersGW, de Crespigny A, Moseley ME. Evalu-ation of early reperfusion and i.v. tPA ther-apy using diffusion- and perfusion-weighted MRI. Neurology 1999; 52:1792–1798.
41. Nagesh V, Welch KM, Windham JP, etal. Time course of ADCw changes inischemic stroke: beyond the human eye!Stroke 1998; 29:1778–1782.
42. Villringer A, Rosen BR, Belliveau JW, etal. Dynamic imaging with lanthanidechelates in normal brain: contrast due tomagnetic susceptibility effects. MagnReson Med 1988; 6:164–174.
43. Rosen BR, Belliveau JW, Aronen HJ, et al.Susceptibility contrast imaging of cere-bral blood volume: human experience.Magn Reson Med 1991; 22: 293–299.
44. Ostergaard L, Sorensen AG, Kwong KK,
Weisskoff RM, Gyldensted C, Rosen BR.High resolution measurement of cere-bral blood flow using intravasculartracer bolus passages. II. Experimentalcomparison and preliminary results.Magn Reson Med 1996; 36:726–736.
45. Ostergaard L, Weisskoff RM, Chesler DA,Gyldensted C, Rosen BR. High resolu-tion measurement of cerebral blood flowusing intravascular tracer bolus pas-sages. I. Mathematical approach and sta-tistical analysis. Magn Reson Med 1996;36:715–725.
46. Rosen BR, Belliveau JW, Vevea JM, BradyTJ. Perfusion imaging with NMR con-trast agents. Magn Reson Med 1990; 14:249–265.
47. Rosen BR, Belliveau JW, Buchbinder BR,et al. Contrast agents and cerebral he-modynamics. Magn Reson Med 1991;19:285–292.
48. Quast M, Huang N, Hillman G, Kent T.The evolution of acute stroke recordedby multimodal magnetic resonance im-aging. Magn Reson Imaging 1993; 11:465–471.
49. Baird A, Benfield A, Schlaug G, et al.Enlargement of human cerebral isch-emic volumes measured by diffusion-weighted magnetic resonance imaging.Ann Neurol 1997; 41:581–589.
50. Rordoff G, Koroshetz W, Copen W, et al.Regional ischemia and ischemic injuryin patients with acute middle cerebralartery stroke as defined by early diffu-sion-weighted and perfusion-weightedMRI. Stroke 1998; 29:939–943.
51. Barber P, Darby D, Desmond P, et al.Prediction of stroke outcome withechoplanar perfusion- and diffusion-weighted MRI. Neurology 1998; 51:418–456.
54. Minematsu K, Fisher M, Li L, et al. Ef-fects of novel NMDA antagonist on ex-perimental stroke rapidly and quantita-tively assessed by diffusion-weightedMRI. Neurology 1993; 43:397–403.
55. Hasegawa Y, Fisher M, Latour L, Dard-zinski B, Sotak C. MRI diffusion map-ping of reversible and irreversible isch-emic injury in focal brain ischemia.Neurology 1994; 44:1484–1490.
56. Muller TB, Haraldseth O, Jones RA, et al.Combined perfusion and diffusion-weighted magnetic resonance imagingin a rat model of reversible middle cere-bral artery occlusion. Stroke 1995; 26:451–457.
57. Minematsu K, Li L, Sotak C, Davis M,Fisher M. Reversible focal ischemic in-jury demonstrated by diffusion-weighted magnetic resonance imaging.Stroke 1992; 23:1304–1310.
58. Dardzinski B, Sotak C, Fisher M, Hase-gawa Y, LI L, Minematsu K. Apparentdiffusion coefficient mapping of ex-
Volume 217 z Number 2 Diffusion-weighted MR Imaging of the Brain z 343
59. Marks MP, De Crespigny A, Lentz D, En-zmann DR, Albers GW, Moseley ME.Acute and chronic stroke: navigatedspin-echo diffusion-weighted MR imag-ing. Radiology 1996; 199:403–408.
60. Gonzalez RG, Schaefer PW, BuonannoFS, et al. Diffusion-weighted MR imag-ing: diagnostic accuracy in patients im-aged within 6 hours of stroke symptomonset. Radiology 1999; 210:155–162.
61. Mohr J, Biller J, Hial S, et al. Magneticresonance versus computed tomographicimaging in acute stroke. Stroke 1995; 26:807–812.
62. Bryan R, Levy L, Whitlow W, Killian J,Preziosi T, Rosario J. Diagnosis of acutecerebral infarction: comparison of CTand MR imaging. AJNR Am J Neurora-diol 1991; 12:611–620.
63. Lovblad K, Laubach H, Baird A, et al.Clinical experience with diffusion-weighted MR in patients with acutestroke. AJNR Am J Neuroradiol 1998; 19:1061–1066.
64. Singer M, Chong J, Lu D, SchonewilleW, Tuhrim S, Atlas S. Diffusion-weighted MRI in acute subcortical in-farction. Stroke 1998; 29:133–136.
65. Ay H, Buonanno FS, Rordorf G, et al.Normal diffusion-weighted MRI duringstroke-like deficits. Neurology 1999; 52:1784–1792.
66. Lovblad K, Baird A, Schlaug G, et al. Isch-emic lesion volumes in acute stroke bydiffusion-weighted magnetic resonanceimaging correlate with clinical outcome.Ann Neurol 1997; 42:164–170.
67. Tong D, Yenari M, Albers G, O’Brien M,Marks M, Moseley M. Correlation of per-fusion- and diffusion-weighted MRI withNIHSS score in acute (,6.5 hour) ischemicstroke. Neurology 1998; 50:864–870.
68. van Everdingen KJ, van der Grond J,Kapelle LJ, Ramos LM, Mali WP. Diffu-sion-weighted magnetic resonance im-aging in acute stroke. Stroke 1998; 29:1783–1790.
69. Tuor U, Kozlowski P, Bigio MD,Ramjiawan B, Su S, Malisza K. Diffusion-and T2-weighted increases in magneticresonance images of immature brainduring hypoxia-ischemia: transient re-versal post hypoxia. Neurology 1998;150:321–328.
70. D’Arceuil HE, de Crespigny AJ, Rother J,et al. Diffusion and perfusion magneticresonance imaging of the evolution ofhypoxic ischemic encephalopathy inthe neonatal rabbit. J Magn Reson Imag-ing 1998; 8:820–828.
71. Cowan FM, Pennock JM, Hanrahan JD,Manji KP, Edwards AD. Early detectionof cerebral infarction and hypoxic isch-emic encephalopathy in neonates usingdiffusion-weighted magnetic resonanceimaging. Neuropediatrics 1994; 25:172–175.
72. Phillips MD, Zimmerman RA. Diffusionimaging in pediatric hypoxic ischemiainjury. Neuroimaging Clin N Am 1999;9:41–52.
73. Ay H, Buonanno F, Schaefer P, et al.Clinical and diffusion-weighted imaging
characteristics of an identifiable subsetof TIA patients with acute infarction (ab-str). In: CD-ROM book of abstracts: 24thAmerican Heart Association Interna-tional Conference on Stroke and Cere-bral Circulation. Dallas, Tex: AmericanHeart Association, 1999.
74. Kidwell CS, Alger JR, DiSalle F, et al. Dif-fusion MRI in patients with transientischemic attacks. Stroke 1999; 30:1174–1180.
75. Ebisu T, Naruse S, Horikawa Y, et al.Discrimination between different typesof white matter edema with diffusion-weighted MR imaging. J Magn ResonImaging 1993; 3:863–868.
76. Tien R, Felsberg G, Friedman H, BrownM, MacFall J. MR imaging of high-gradecerebral gliomas: value of diffusion-weighted echoplanar pulse sequences.AJR Am J Roentgenol 1994; 162:671–677.
77. Schaefer P, Buonanno F, Gonzalez R,Schwamm L. Diffusion weighted imag-ing discriminates between cytotoxic andvasogenic edema in a patient witheclampsia. Stroke 1997; 28:1–4.
78. Schwartz R, Mulkern R, Gudbjartsson H,Jolesz F. Diffusion-weighted MR imag-ing in hypertensive encephalopathy:clues to pathogenesis. AJNR Am J Neu-roradiol 1998; 19:859–862.
79. Tsuruda J, Chew W, Moseley M, Nor-man D. Diffusion-weighted MR imagingof the brain: value of differentiating be-tween extra-axial cysts and epidermoidtumors. AJNR Am J Neuroradiol 1990;11:925–931.
80. Maeda M, Kawamura Y, Tamagawa Y, etal. Intravoxel incoherent motion (IVIM)MRI in intracranial, extra-axial tumorsand cysts. J Comput Assist Tomogr 1992;16:514–518.
81. Brunberg J, Chenevert T, McKeever P, etal. In vivo MR determination of waterdiffusion coefficients and diffusion an-isotropy: correlation with structural al-teration in gliomas of the cerebral hemi-spheres. AJNR Am J Neuroradiol 1995;16:361–371.
82. Eis M, Els T, Hoehn-Berlage M, Houss-man K. Quantitative MR imaging of ce-rebral tumor and edema. Acta Neurochir1994; (suppl):344–346.
83. Els T, Eis M, Hoehn-Berlage M, Hoss-mann K. Diffusion-weighted imaging ofexperimental brain tumors in rats.MAGMA 1995; 3:13–20.
84. Aronen H, Cohen M, Belliveau J,Fordham J, Rosen B. Ultrafast imaging ofbrain tumors. Top Magn Reson Imaging1993; 5:14–24.
85. Eis M, Els T, Hoehn-Berlage M. High res-olution quantitative relaxation and dif-fusion MRI of three different experimen-tal brain tumors in rats. Magn ResonMed 1995; 34:835–844.
86. Ostergaard L, Hochberg F, Rabinov J, etal. Early changes measured by magneticresonance imaging in cerebral bloodflow, blood volume and blood brain bar-rier permeability following dexametha-sone treatment in patients with braintumors. J Neurosurg 1999; 90:300–305.
87. Kim Y, Chang K, Kim H, Seong S, Kim Y,Han M. Brain abscess and necrotic orcystic brain tumor: discrimination with
signal intensity on diffusion-weightedMR imaging. AJR Am J Roentgenol 1998;171:1487–1490.
88. Ebisu T, Tanaka C, Umeda M, et al. Dis-crimination of brain abscess from ne-crotic or cystic tumors by diffusion-weighted echo planar imaging. MagnReson Imaging 1996; 14:1113–1116.
89. Schaefer P, Wang B, Gonzalez R. Echoplanar diffusion weighted imaging inpyogenic infections (abstr). In: Proceed-ings of the 35th Annual Meeting of theAmerican Society of Neuroradiology.Oak Brook, Ill: American Society of Neu-roradiology, 1997; 103.
90. Bahn M, Kido D, Lin W, Pearlman A.Brain magnetic resonance diffusion ab-normalities in Creutzfeldt-Jakob disease.Arch Neurol 1997; 54:1411–1415.
91. Schaefer P, Grant E, Wang B, GonzalezR. Echo planar diffusion weighted imag-ing (DWI) in Creutzfeldt-Jakob disease(CJD) (abstr). In: Proceedings of the 37thAnnual Meeting of the American Societyof Neuroradiology. Oak Brook, Ill: Amer-ican Society of Neuroradiology, 1999;161.
92. Demaerel P, Heiner L, Robberecht W,Sciot R, Wilms G. Diffusion-weightedMRI in sporadic Creutzfeldt-Jakob dis-ease. Neurology 1999; 52:205–208.
93. Brown P, Cathala F, Castaigne P, Gaj-dusek DC. Creutzfeldt-Jakob disease:clinical analysis of a consecutive seriesof 230 neuropathologically verifiedcases. Ann Neurol 1986; 20:597–602.
94. Johnson R, Gibbs C. Creutzfeldt-Jakobdisease and related transmissible spongi-form encephalopathies. N Engl J Med1998; 339:1994–2004.
95. Finkenstaedt M, Szudra A, Zerr I, et al.MR imaging of Creutzfeldt-Jakob dis-ease. Radiology 1996; 199:793–798.
96. Hanstock C, Faden A, Bendall M, Vink R.Diffusion-weighted imaging differenti-ates ischemic tissue from traumatizedtissue. Stroke 1994; 25:843–848.
97. Yuan X, Prough D, Smith T, DeWitt D.The effects of traumatic brain injury inregional cerebral blood flow in rats.J Neurotrauma 1988; 5:289–301.
98. Ito J, Marmarou A, Barzo P, Fatouros P,Corwin F. Characterization of edema bydiffusion-weighted imaging in experi-mental traumatic brain injury. J Neuro-surg 1996; 84:97–103.
99. Barzo P, Marmarou A, Fatouros P, Ha-yasaki K, Corwin F. Contribution of va-sogenic and cellular edema to traumaticbrain swelling measured by diffusion-weighted imaging. J Neurosurg 1997; 87:900–907.
100. Schaefer P, Tievsky A, Gonzalez R. Echoplanar (EP) diffusion weighted imaging(dwi) in diffuse axonal injury (dai) (ab-str). In: Proceedings of the 36th AnnualMeeting of the American Society of Neu-roradiology. Oak Brook, Ill: AmericanSociety of Neuroradiology, 1998; 216.
101. Atlas SW, Dubois P, Singer MB, Lu D.Diffusion measurements in intracranialhematomas: implications for MR imag-ing of acute stroke. AJNR Am J Neurora-diol 2000; 21:1190–1194.
102. Heide A, Richards T, Alvord E, Peterson J,Rose L. Diffusion imaging of experimentalallergic encephalomyelitis. Magn ResonMed 1993; 29:478–484.
344 z Radiology z November 2000 Schaefer et al
103. Verhoye MR, ’s-Gravenmade EJ, RamanER, Van Reempts J, Van der Linden A. Invivo noninvasive determination of ab-normal water diffusion in the rat brainstudied in and animal model for multi-ple sclerosis by diffusion-weighted NMRimaging. Magn Reson Imaging 1996; 14:521–532.
104. Horsfield M, Lai M, Webb S, et al. Ap-parent diffusion coefficients in benign
and secondary progressive multiple scle-rosis by nuclear magnetic resonance.Magn Reson Med 1996; 36:393–400.
105. Christiansen P, Gideon P, Thomsen C,Stubgaard M, Henricksen O, Larsson H.Increased water self-diffusion in chronicplaques and in apparently normal whitematter in patients with multiple sclerosis.Acta Neurol Scand 1993; 87:195–199.
106. Larsson H, Thomsen C, Frederiksen J,
Stubgaard M, Henriksen O. In vivo mag-netic resonance diffusion measurementin the brain of patients with multiple scle-rosis. Magn Reson Imaging 1992; 10:7–12.
107. Allen IV, McKeown SR. A histological,histochemical and biochemical study ofthe macroscopically normal white mat-ter in multiple sclerosis. J Neurol Sci1979; 41:81–91.
Volume 217 z Number 2 Diffusion-weighted MR Imaging of the Brain z 345
