Sensory Motor Cortex: Correlation of Presurgical Mapping ... · motor cortex wasperformed intwo...
Transcript of Sensory Motor Cortex: Correlation of Presurgical Mapping ... · motor cortex wasperformed intwo...
Clifford R. Jack, Jr. MD #{149}Richard M. Thompson, MD #{149}R. Kim Butts, PhDFrank W. Sharbrough, MD #{149}Patrick J. Kelly, MD #{149}Dennis P. Hanson, BSStephen J. Riederer, PhD #{149}Richard L. Ehman, MD #{149}Nicholas J. Hangiandreou, PhDGregory D. Cascino, MD
Sensory Motor Cortex: Correlation ofPresurgical Mapping with Functional MRImaging and Invasive Cortical Mapping’
85
Neuroradiology
PURPOSE: To describe a clinicallyuseful application of functional mag-netic resonance (MR) imaging-pre-surgical mapping of the sensory mo-tor cortex-and to validate the resultswith established physiologic tech-niques.
MATERIALS AND METHODS: Func-tional MR mapping of the sensorymotor cortex was performed in twowomen, aged 24 and 38 years. Bothhad intractable, simple partial motorseizures due to tumors located in ornear the sensory motor cortex. Theysubsequently underwent invasivecortical mapping-direct cortical
stimulation and/or sensory-evoked-potential recording-to localize theaffected sensory motor area prior totumor resection.
RESULTS: In both patients, the func-tional MR study demonstrated taskactivation of the sensory motor cor-tex. In both cases, results of corticalfunctional mapping with invasivetechniques matched those obtainedwith functional MR imaging.
CONCLUSION: Presurgical mappingof the sensory motor cortex is a po-tentially useful clinical application offunctional MR imaging.
Index terms: Blood, flow dynamics #{149}Brain,
blood flow, 10.919 #{149}Brain, function, 10.919
Epilepsy #{149}Magnetic resonance (MR), vascular
studies
Radiology 1994; 190:85-92
I From the Departments of Diagnostic Radiol-
ogy (C.R.J., R.M.T., R.K.B., S.J.R., R.LE., N.J.H.),Neurology (F.W.S., G.D.C.), Neurosurgery(P.1K.), and Biomedical Imaging Resource(D.P.H.), Mayo Clinic and Foundation, 200 FirstSt SW, Rochester, MN 55905. Received June 2,1993; revision requested July 27; revision re-ceived August 23; accepted September 7. Sup-ported in part by Public Health Service grantsno. ROl NS28374, ROl CA37993, and HL07269-15. Address reprint requests to CR1.
� RSNA, 1994
I F surgical resection of a lesion lo-
cated in or near functionally es-
sential cortex is considered, then lo-
calization of functional areas relative
to the surgical target (the lesion) must
be ascertained to avoid a postopera-
tive neurologic deficit. This is particu-
larly critical in patients whose only
symptom is epilepsy. Such patients
tend to have benign or indolent non-life-threatening lesions, and a postop-
erative neurologic deficit may repre-
sent an unacceptable surgical risk.Traditionally, presurgical functional
localization has been accomplished
with invasive means: (a) direct corticalstimulation either intraoperatively in
the awake patient or postoperatively
after subdural grid placement or
(b) sensory-evoked-potential studies
after operative grid placement.
It has been shown recently thatfunctional magnetic resonance (MR)
imaging with blood oxygen level-
dependent (BOLD) contrast is capableof noninvasively depicting primary
sensory areas including the sensory
motor cortex (1-9). However, studies
to date have used volunteers andthere has been no direct validation of
the physiologic truth of functional
localization with MR imaging. In thisarticle we describe a clinically useful
application of functional MR map-
ping of the sensory motor cortex forsurgical planning. In addition, we
compare the results of localization
with functional MR imaging with theaccepted criterion standard of inva-
sive cortical mapping.
MATERIALS AND METHODS
Case 1
The patient was a 24-year-old womanwho had been having simple partial motorseizures since she was 7’/2 years old. When
she was evaluated at our institution, shewas having two to five seizures per month
despite treatment with an optimal antiepi-leptic medical regimen. Her seizures be-
gan with an aura of right facial numbness
and a painful sensation in the throat fol-lowed by choking and speech arrest. Post-
ictally she was dysanthric with a right
facial droop. Seizures were not accompa-
nied by loss of consciousness. At age 19years, a computed tomography (CT) scanwas reportedly negative, and at age 23
years an MR study demonstrated a leftfrontoparietal lesion (Fig 1). She was
evaluated at our institution as a surgical
candidate and was admitted for pro-
longed inpatient video-electroencephalo-graphic monitoring with withdrawal of
medication. Ictal onset was determined to
be of left frontal origin. On the basis of
both clinical and electroencephalographiccriteria, the site of seizure onset was in or
near the left sensory motor cortex. Preop-erative testing also included speech, lan-guage, and neuropsychologic evaluation,the results of which were within normallimits. Physical examination revealed mild
right facial weakness, but findings were
otherwise negative. Cerebral angiography
with amytal testing demonstrated lefthemispheric speech dominance.
A functional MR examination with taskactivation of the sensorimotor cortex wasperformed with use of an approach similarto that described by Connelly et al (8). Aseries of short Ti-weighted acquisitions inthe axial and oblique planes were ob-
tamed to localize the tumor. From theseTi-weighted images, several anatomicplanes of interest were selected in whichto perform functional MR imaging. The
functional MR sequence consisted of ob-taming 20 consecutive 3D SPGR images ata single section location with the follow-ing parameters: echo time, 60 msec; repeti-tion time, 80 msec; flip angle, 40#{176};field ofview, 24 cm; section thickness, 4 mm; onesignal averaged; a 64 x 64 matrix; and 5.i
seconds per image. These images wereobtained on a i.5-T system (Signa; GEMedical Systems, Milwaukee, Wis) with
standard hardware and software (the only
software modification was that permitting
use of the coarse acquisition matrix). Dur-
ing this 20-image acquisition, the patient
Abbreviations: BOLD = blood oxygen leveldependent, 3D SPGR = three-dimensionalspoiled gradient-recalled acquisition in thesteady state.
a. b.
86 #{149}Radiology January 1994
Figure 1. Patient 1. Diagnostic MR images. (a) Axial and (b) reformatted coronal images from a three-dimensional spoiled gradient-recalled
acquisition in the steady state (3D SPGR) sequence. These images demonstrate a peripherally located left frontopanietal tumor, with a complex
relationship to the local normal sulcal anatomy. The inferior border of the tumor is at the brain surface and erodes the adjacent inner table, in-dicating its indolent nature.
alternated between rest and a voluntaryactivation task, which consisted of bilat-eral fingers to thumb opposition and con-traction and relaxation of the lip andlower face muscles to activate the handand lip-lower face portion of the sensorymotor homunculus. The lip-lower facetask was performed with no jaw motion(ie, only facial muscles around the mouth
were moved) to minimize the possibilityof head movement. Image processing con-sisted of simple image subtraction. The 20images were partitioned into four clustersof five inactive, five active, five inactive,and five active images. The first image ofeach cluster was eliminated to produce asteady magnetization state for the firstcluster of images and a physiologic steadystate for the remaining three clusters. Theactive images were then added togetherand subtracted from the sum of the mac-tive images.
To anatomically link areas of activationon functional MR images to cortical sun-face anatomy, a volume rendering of thebrain surface was performed with use of
software (Analyze; Biomedical ImagingResource, Mayo Clinic and Foundation,Rochester, Minn) (10). The brain was seg-mented from overlying structures bymeans of a series of mathematical erode,connect, and dilate morphology opera-lions. The segmented MR imaging datawere then loaded into a program that si-
multaneously displayed the volume-ren-dered image of the brain surface againstthe cross-sectional plane in which thefunctional MR images were obtained.
The patient subsequently underwentsurgical implantation of a series of subdu-ral recording strips. A small craniotomywas performed over the tumor, which had
been localized stereotaxically. Three verti-cally oriented strips were placed above thetumor, and one horizontal T strip wasplaced below the tumor. Each recordingstrip consisted of eight circular stainlesssteel electrode contacts that were 4 mm in
diameter, iO mm in center-to-center dis-tance, and embedded in a 75-mm-longSilastic (Dow Corning, Midland, Mich)sheath. Eight wire leads emerged fromone end of the vertically placed strips andfrom the middle of the horizontally placedT strip. These leads passed through thecraniotomy and were connected to theexternal circuitry. A cortical stimulationstudy was performed the following day.Current was passed between pairs of elec-trodes in a systematic fashion. The rela-tionship between specific pairs of elec-trodes and elicited sensory and/or motorfunctions of the hand, foot, and face wasrecorded. Also recorded were those stimu-lation pains that reproduced the patient’s
aura.A thin-section CT study was then per-
formed with the subdural recording stripsin place. The CT study was co-registered
with a 3D SPGR MR sequence obtainedpreoperatively by using a surface-match-ing algorithm (ii). The position of each
individual electrode contact was extracted
from the registered CT data and was
stored as a separate object in an objectmap. The brain and the tumor were seg-mented from the 3D SPGR images andwere stored as separate objects in the ob-ject map. A volume-rendered image of thebrain surface with the tumor and each in-dividual electrode contact rendered as
separate objects was generated by using a24-bit color compositing multiple-objectvolume-rendering technique (10).
The patient then underwent surgicalremoval of the tumor. Prior to tumor re-moval, intraoperative sensory evoked p0-tential tests were performed. Pathologic
examination revealed a grade 2 oligoden-droglioma. Postoperatively, findings at the
patient’s neurologic examination were un-
changed.
Case 2
The second patient was a 38-year-oldwoman who was well until she experi-enced several secondarily generalizedtonic-clonic seizures at age 30 years. MR
imaging performed at that time demon-
strated a right frontal intraaxial tumor (Fig
2). When she was seen at our institution,the patient was experiencing approxi-mately two seizures per week despite re-ceiving optimal medical therapy. Theseseizures began with an aura of numbness
in the left hand or hemibody, and ap-
proximately iO% of the auras proceeded
to become focal motor seizures involvingthe left upper and lower extremity withspeech arrest but with no loss of con-
sciousness. Preoperatively, single-sectionfunctional MR imaging was performed in
several anatomic planes of interest. Sepa-
rate hand and foot activation tasks werepositioned as outlined for patient 1. Pro-
longed scalp-recorded inpatient video-electroencephalographic monitoring re-vealed right frontal seizure onset. Duringthe surgical procedure, recording stripswere placed over the paracentral cortexand sensory-evoked-potential tests withmedian nerve stimulation were performed
to document the position of the centralsulcus in relation to the tumor. The tumor
Figure 2. Patient 2. Diagnostic MR images. Axial (a) T2-weighted and (b) Ti-weighted im-ages reveal an infiltrative right frontal tumor.
Volume 190 #{149}Number I Radiology #{149}87
was removed and identified as a grade 3
oligodendroglioma. The patient experi-enced no motor deficit as a result of sur-
gery.
Case 1
RESULTS
The functional MR images of thispatient performing the task describeddemonstrated clear-cut cortical activa-tion centered between two verticallyoriented gyri. We expected these to bethe pre- and postcentral gyri, with theactivation centered in the central sul-cus. On the basis of the findings in the
functional MR study, we concludedthat the tumor straddled the inferiorportion of the central sulcus (Fig 3).The functional task administeredshould have activated both the handand the lip-lower face portions of thesensorimotor homunculus (Fig 4) (12).
In Figure 3b the majority of the acti-vated area lies well above the tumor;the lowest portion of the activatedstrip is located at the uppermost bor-der of the tumor. On the basis of find-ings in the images in Figure 3, we pre-dicted that resection of the tumorwould involve the lip-lower face por-lion of the sensorimotor homunculusbut not the more cephalic hand area.As a rule, surgical damage to the handor foot portion of the homunculusproduces a severe functional deficit
and is to be avoided, while damage tothe face-lip portion will produce aclinically negligible deficit.
Figure 5 is a schematic drawing of
the position of the four subdural re-cording strips in relation to the tumor
and of the results of the cortical
stimulation studies. Stimulation of thevertical strips revealed the expectedorientation of the sensory motor ho-munculus of Penfleld and Rasmussen(12). From the stimulation studies itwas deduced that strip A was located
over the motor cortex, strip B over thesensory cortex, and strip C posterior
to the sensory cortex (Fig 5). The cen-tral sulcus was therefore located be-tween strips A and B. Intraoperativesensory-evoked-potential recordingwith median nerve stimulation dem-onstrated maximum response at elec-trode 4 on strip B, thereby also con-firming that strip B was over theprimary sensory cortex. Stimulationbetween the lower electrodes onstrips A and B, as well as between
electrodes 4 and 5 on strip D, repro-duced the patient’s aura, a sensationin the throat. Intraoperatively, onlythe most inferior contacts of the threevertically oriented strips and contacts4-6 of strip D were visible (Fig 6). Theremaining contacts were hidden be-neath the intact skull. Intraoperativephotographs (Fig 6) document thatthe portion of the tumor at the brainsurface was located between and infe-rior to contact 8 of strips A and B andabove contacts 4 and 5 of strip D. Be-cause of the limited exposure of the
brain surface provided by the smallcraniotomy and because of the factthat the tumor had distorted normallocal anatomic landmarks, an ana-tomic central sulcus could not beidentified by inspecting the operativefield. Because intraoperative inspec-tion did not provide the desired cor-
relation between the invasive stimula-tion studies and functional MRmapping, the following strategy waspursued.
A volume rendering of the brainsurface was generated with the tumorand electrode contacts rendered asseparate objects (Fig 7). This render-ing demonstrated the relationship ofindividual electrode contacts to boththe tumor and the cortical topogra-
phy that could not be appreciatedintraoperatively. This is a useful tech-nique, because it enables correlationbetween the results of direct corticalstimulation and functional MR imag-ing. Visual comparison of the volume-rendered brain surface images in Fig-
ures 3 and 7 confirmed the impressionformed from the functional MR studythat the tumor straddled the func-tional sensorimotor strip.
Case 2
In patient 2, images of only onetask (opposition of repetitive fingers
to thumb) demonstrated convincingfunctional activation (Fig 8). Failedimages were most likely a result ofpatient motion. As this patient haddifficulty with head motion, the func-
tional image (Fig 8) was generatedwith a single inactive-active cycle.Because of artifacts in this image,color mapping of functional activa-tion onto a Ti-weighted anatomictemplate was not effective. Instead, alinked-cursor display was used toidentify common pixels in the Ti-
weighted and functional images (10).Figure 8b demonstrates a curvilineararea of activation centered on a sulcus(the central sulcus) that is posterior tothe tumor. We inferred from this thatthe tumor extended posteriorly to butdid not directly involve the primarymotor strip. On the basis of findingsin this image, we predicted that resec-lion of the tumor would spare theprimary motor cortex.
At surgery, the craniotomy was notextended beyond the posterior mar-gin of the tumor, and therefore therelationship between the tumor andthe cortical topography posterior to itcould not be appreciated visually.However, intraoperative sensory-evoked-potential recording was per-formed by inserting subdural recording
strips beneath the posterior margin of
the craniotomy flap. Findings at leftmedian nerve stimulation demon-strated that the central sulcus waslocated approximately 2 cm posteriorto the posterior margin of the tumor,thus confirming the impressionformed at functional MR imaging.
10
88 #{149}Radiology January 1994
Figure 3. Patient 1. Functional MR images. ++ct�e I +ct��e � �+ct� � act�e I ‘+ct�+e octi�+ �+ct�+e
(a, b) Four-panel collages. The upper leftpanel is a Ti-weighted anatomic template of 20
the cross section in which the corresponding
functional MR imaging sequence was ob- � � �tamed (a small arrow indicates the position �
of the tumor). The upper right panel repre- � � �sents the functional activation image formed � �, I 10
by the addition and subtraction process out- �lined herein. The lower right panel is created I
by assigning a color map to the functionalimage and then fusing that image with the � � I
Ti-weighted anatomic template. The lower 20 30 40 10 20 30 40
left panel represents a volume rendering of Image �umber Image Number
the brain surface with the tumor rendered as C. d.a separate green object and with the plane inwhich the functional MR image was obtained indicated by a line (as in a) or a shaded gray planar surface (as in b). Part a was obtained axially
through the top of the tumor. Activation of the sensory motor cortex bilaterally is seen in the top right functional activation image. The rela-tionship of the activated sensorimotor cortex in the patient’s left hemisphere to the top of the tumor is illustrated in the fused image in the bot-torn right panel. Part b is an oblique section with anterior (ANT), posterior (POST), superior (SUP), and inferior (INF) labeled for orientation inthe Ti-weighted image in the top left panel. The curvilinear area of activation (top and lower right panels) precisely follows the contour of the
central sulcus. The most inferior portion of this activated strip should represent the lip-lower face portion of the homunculus, whereas themore superior portion should represent the hand portion. The tumor is centered on and elevates the most inferior portion of the central sulcus,the top of which is indicated by a large arrow in the lower left images. (c, d) Time course plots of signal intensity in a manually defined regionof interest. Part c is from a region of interest in the left sensorimotor area in a, and part d is from the sensorimotor area in b. Both c and d dem-onstrate a cyclic change in signal intensity, which follows the periodicity of the activation task.
DISCUSSION
Functional activation of the cortexwith MR imaging was first demon-strated by means of a contrast bolus-tracking technique (i3). It has been
subsequently shown that this phe-nomenon can be visualized without
exogenously administered contrastmaterial (1-9). Fox and Raichle (14)demonstrated with positron emissiontomography that appropriately de-signed stimulation paradigms willproduce a local change in a number ofphysiologic parameters from baseline,in appropriate areas of the cortex.
Most relevent to this type of imagingare the facts that cerebral perfusionwill increase locally and will do so inexcess of the oxygen metabolic rate.This results in both an increase in tis-sue perfusion and a paradoxical netdecrease in concentration of deoxyhe-moglobin in the capillary and venousbed of an activated area of cerebralcortex. Oxygenated hemoglobin isdiamagnetic whereas deoxyhemoglobin
is paramagnetic (15,16). The paramag-netic properties of deoxyhemoglobincreate local field inhomogeneities,which decrease intravoxel spin coher-ence in the vicinity of blood vessels
and thereby decrease the signal inten-sity on T2- or T2*�weighted MR images(15-20). Ogawa et al (17,18) suggested
that the state of blood oxygenationmay be a useful contrast parameter atMR imaging, and they coined the ac-ronym BOLD contrast. Two possiblemechanisms may contribute to the in-
creased signal intensity seen in func-tionally activated cortical areas: Oneis the BOLD effect. The other is an in-
creased flow of unsaturated spins into
the imaging section which cycles inphase with the periodicity of the acti-vation task. The precise interrelation-ship between these two mechanisms
I
Figure 4. Motor homunculus. Pictorial representation of the loca-tion and relative extent of the different portions of the motor ho-munculus on the precentral gyrus. Analogous sensory areas arerepresented on the postcentral gyros. (Reprinted, with permission,from reference i2.)
sup.
A
A.\�F.
(r-�-Th10 E�o N
ISO
1150�
Aura
Throat �
-Little Finger
Seas
- ThumbSeas
C
0 E ‘-‘
ONIS 2
0000
00
- Little FingerMot
Thumb
Mot
Aura
Throat
0[�OOO�JOO�JD
IN!2.
Figure 5. Patient I. Subdural strips schematic of the position of the four subdural recordstrips (A, B, C, D) in relation to the tumor: anterior (ANT), superior (SUP), posterior (POST),
and inferior (INF). MOT = motor response, SENS = sensory response. The central sulcus lies
between strips in a and b. Results match the Penfield homunculus seen in Figure 4.
Volume 190 #{149}Number 1 Radiology #{149}89
(��Th00OTTO
JIM
A0 �
NO
DO
00
Aura
Throat
and observed activation-inducedchange in MR signal intensity is notyet clear. Use of a small flip angle and
a phase-encoding scheme in whichthe central lines of K space are en-coded first should decrease the effectof inflow enhancement; however, nei-
ther of these were employed in thesetwo patients (5,7).
Most functional MR imaging workto date has focused on the BOLD ef-fect. BOLD contrast results from a netdecrease in concentration of deoxyhe-moglobin in the capillary and venous
system during task activation. Ideally,
the accompanying changes in signalintensity at MR imaging serve as aprecise marker of those areas of cortexthat have undergone task activation.However, it is theoretically possiblethat the net influx of oxyhemoglobininto the venous system surroundingan area of activated cortex may bewashed downstream, away from the
specific area of task-activated cortex.Conceivably, alterations in signal in-tensity associated with task activation
could be present in large superficialveins projecting over areas of cerebralcortex that were not involved in theactivation process. If this were thecase, then BOLD functional MR imag-ing would be far less specific and lessuseful for cortical mapping than if thealterations in signal intensity re-mained confined to the capillary bedor to small veins in the sulcus physi-cally adjacent to the areas of cortexthat were directly activated. It hasrecently been shown that the weight-ing of BOLD contrast with respect tovessel size varies as a function of bothstrength of static magnetic field andtype of echo (20,21). Imaging at highfield strength (4 T) and with a spin-echo technique produces greater con-trast weighting for small vessels (ie, 3�im or capillary level) than does imag-ing at lower field strength (1.5 T) orwith a gradient-recalled echo. How-ever, the observed activation-inducedchange in signal intensity is consider-ably smaller with spin-echo than with
POST. gradient-recalled-echo techniques.The inherently small contrast-to-noiseratio makes functional imaging withspin-echo techniques particularlychallenging at i.5 T. The implicationsfor clinical applications of functionalMR imaging are obvious. Few siteswill ever have 4-T whole-body imag-ers. Therefore, clinical applicationswill likely be investigated at 1.5-2.0 Twith use of techniques (ie, gradientecho) that favor visualization of veinsrather than capillaries. This is a prob-lem, however, only if the functionalresolution of the MR technique is toocoarse for the clinical task. In viewingthe functional images of patient 1,particularly those in Figure 3b, it oc-curred to us that the activation signalintensity might partially or com-pletely represent the large central sul-cus vein seen on the patient’s preop-
erative angiogram. However, whenthe morphology of the central sulcus
vein (Fig 9) is compared with that ofthe activation signal intensity in Fig-ure 3, a clear difference can be seen.We therefore conclude that while theactivation signal intensity in Figure 3
a. b.
Figure 7. Patient 1. (a) In CT scan with recording strips in place, the individual electrode
contacts in strips A, B, and C can be seen over the left vertex. (b) Lateral view with the
patient’s head rotated slightly downward. The brain surface is rendered as a partially trans-parent object. The tumor is rendered in green, and that portion of the tumor that is at thebrain surface appears brighter than the portion of the tumor that is deep to the brain surface.The eight individual electrode contacts of each of the four subdural strips are rendered asseparate red objects. Strips A, B, C, and D are labeled as they are in Figure 5 and Figure 6. The
central sulcus is shown by means of cortical stimulation to lie between strips A and B. There-fore, the tumor straddles the inferior portion of the central sulcus. Correlation of this imagewith that in Figure 3 confirmed the relationship between the localization of function on thecortical surface with functional MR imaging and that obtained with invasive recording.
90 #{149}Radiology January 1994
Figure 6. Patient 1. Intraoperative photographs. After stereotaxic localization, a small burr hole was drilled directly over the tumor. (a) Corti-cal surface prior to placement of the recording strips and (b) strips on the cortical surface for comparison with Figure 5. The strips are labeled A,
B, C, D as in Figure 5. Note only the most inferior electrode of the three vertical strips and electrodes 4, 5, and 6 of the horizontal strip are vis-
ible in the operative field, because the strips have been slipped beneath the dura beyond the craniotomy. For purposes of orientation, the
patient’s nose is on the reader’s left, the top of the image is the superior part of the head, the electrodes in strip D overlie the sylvian fissureand the inferior pant of the frontoparietal operculum. (c) The cortex after the tumor has been removed. The most inferior portion of the tumorwas entered and the surgical plane extended superiorly beneath the brain surface. From this picture it is difficult to appreciate the entire extentof tumor removal, as the superior portion of the operative bed is concealed beneath undisturbed cerebral cortex. In comparison with Figure 9,note the tumor lies in the base of a U-shaped vein that extends in a cephalic direction, beneath the superior portion of the craniotomy flap.
may in part represent the effects ofvolume averaging of the superficiallylocated central sulcus vein, it is morelikely a result of signal in small veinsin the deeper banks of the central sul-cus that lie directly adjacent to theactivated cortex.
The method employed here to gen-erate functional MR images involvedsimple addition and subtraction of thetime-course images. A drawback ofthis approach is that motion (brain orblood pulsation or bulk head motion)during data acquisition will tend tocorrupt what is already a procedurewith low signal-to-noise ratio. Thenmlike high-intensity artifacts at thebrain surface in Figures 3b and 8bdemonstrate this problem. Pulsatilemotion of cerebrospinal fluid-whichis probably not a major problem inthe images discussed herein becauseof the radio-frequency spoiling andlarge flip angle-will also corrupt
functional images obtained with tech-niques that produce a high signal in-tensity in cerebrospinal fluid. Severalattempts at producing functional MRimages in patient 2 failed to show anyobvious result, probably because ofhead motion during serial image ac-
quisition. More sophisticated imageprocessing procedures have been pro-posed such as t test or z maps de-signed to minimize these artifacts(22). A most promising approach by
Bandettini et al (23) involves pixel-by-pixel thresholding according to the
cross correlation between the shape
of the time-course image data andthat of the stimulus cycle.
Presurgical mapping of the sensori-motor cortex is an appealing clinicalapplication of functional MR imaging.To our knowledge, results of func-
tional MR studies published to datehave not included rigorous physi-ologic correlation between presumedfunctional localization with functionalMR imaging and actual localizationwith established invasive recordingtechniques. Surgery for epilepsy nearfunctionally essential cortical areas isone of the few instances in which di-rect invasive brain mapping is mdi-
cated clinically, thus providing a
unique opportunity for verifying thephysiologic truth of functional MRimaging. Because this procedure isrelatively uncommon (even in centers
with a large volume of epilepsy sur-gery), experience with this correlation
will accumulate slowly on a case-by-case basis. The two cases presented
herein preliminarily support the fidel-ity of functional MR imaging in local-
izing the sensory motor cortex. Theprobable localization of the activation
signal in small cortical veins (versus
capillaries) does not appear to ham-
per this application significantly. Prior
to widespread performance of func-tional MR imaging for this or otherclinical uses, its accuracy must be vali-
dated against that of standard meth-
active
0/2
h’:, .
:� �
� :s� . �
�
Image Number
a. c.
Figure 8. Patient 2. Functional MR imaging study. (a) Coronal scout image with oblique axialplanes cross referenced. (b) Images were obtained in section location 2. At top left is an ob-
liquely oriented Ti-weighted image through the tumor. Posterior (POST), superior (SUP),
and inferior (INF) are indicated for orientation. Functional image at top right, obtainedthrough the same section location as the Ti-weighted map shows a hand movement activa-lion task. A small S-shaped area of activation (arrow at top right) is seen in the posteroinferioraspect of the image, corresponding to hand activation. The middle and lower panel demon-
strate a linked cursor display. The intersection of the cross hairs in the Ti-weighted image con-responds to the same spatial coordinates in the adjacent functional image. These demonstrate
that the S-shaped area of functional activation is centered on a sulcus (the central sulcus) that
is located posterior and inferior to the tumor. On the basis of findings in the functional MR
imaging study, we predicted that the primary motor cortex was located posterior to the tu-mor. (c) Time course plot of signal intensity in region of interest defined manually in the sen-sonimotor area in b. This was a single inactive-active cycle, as the patient had difficulty with
head motion.
b.
Volume 190 #{149}Number 1 Radiology #{149}91
Figure 9. Patient 1. Angiograms. (a) A lucent area (arrow) is seen in the midportion of the plain skull angiogram. (b) The venous phase of the
angiogram demonstrates that this lucency is located precisely in the anteroinfenior aspect of a U-shaped draining cortical vein (arrows). (c) Asubtraction angiogram demonstrates the central cortical vein more clearly (arrows). Proof of the anatomic relationship between the tumor and
this vein is obtained by comparing the shape of this vein in the angiogram with the shape of the draining vein that hugs the inferior aspect of
the tumor in the intraoperative photographs of Figure 6. In turn, by comparing the shape of the vein with the curvilinear shape of the area of
functional activation in Figure 3, one can see that although they are both located in the central sulcus, there is only a moderate physical corre-spondence between the two. The shape of the functional activation (Fig 3) is more complex and follows the curving contour of the portion ofthe central sulcus that lies deep to the surface. In contrast, the large central sulcus vein is less convoluted and lies on the brain surface.
ods of brain mapping. The cases re-
ported herein illustrate a paradigmfor performing these necessary vali-dation studies. Because subdural re-cording grids or strips often lie par-tially under intact skull and may shift
position during surgery, accurate in-traoperative assessment of electrode
position is difficult. Therefore, thefunctional-anatomic link betweenfunctional MR imaging and direct re-cording demonstrated in Figures 3and 7 is particularly useful. #{149}Acknowledgments: The authors thank BrendaMaxwell and Cindy Pfremmer of Mayo Bio-
medical Imaging Resource for manuscript typ-ing.
References1. Ogawa 5, Tank DW, Menon R, et al. In-
trinsic signal changes accompanying sen-sory stimulation: functional brain mappingwith magnetic resonance imaging. Proc
NatI Acad Sci USA 1992; 89:5951-5955.2. Kwong KK, Belliveau JW, Chester DA, et at.
Dynamic magnetic resonance imaging ofhuman brain activity during primary sen-sony stimulation. Proc Natl Acad Sci USA1992; 89:5675-5679.
3. Bandettini PA, Wong EC, Hinks RS, et at.
Time course EPI of human brain functionduring task activation. Magn Reson Med1992; 25:390-397.
4. Turner R, Jezzard P, Wen H, Kwong KK, et
al. Functional mapping of the human vi-suat cortex at 4 and 1.5 Tesla using deoxy-genation contrast EPI. Magn Reson Med
1993; 29:277-279.
5. Frahm J, Bruhn H, Merbloldt K, HanickeW. Dynamic MR imaging of human brainoxygenation during rest and photic stimu-lation. JMRI 1992; 2:501-505.
6. Belliveau JW, Kwong KK, Kennedy DN, etat. Magnetic resonance imaging mappingofbrain function: human visual cortex. In-vest Radiol 1992; 27:S59-565.
7. Menon RS, Ogawa 5, Kim SG, et at. Func-tional brain mapping using magnetic reso-
nance imaging: signal changes accompany-ing visual stimulation. Invest Radiot 1992;27:S47-S53.
92 #{149}Radiology January 1994
8. Connelly A,Jackson GD, Frackowiak RSJ,et al. High-resolution functional mappingof activated human primary cortex using a
clinical magnetic resonance imaging sys-tem. Radiology 1993; 188:125-130.
9. Kim SG, Ashe J, Hendrich K, et al. Func-tional magnetic resonance imaging of mo-ton cortex: hemispheric asymmetry andhandedness. Science 1993; 261:615-617.
10. Robb RA. A software system for intenac-tive and quantitative analysis of biomedicalimages. In: H#{246}hneKH, Fuchs H, Pizer SM.3D imaging in medicine. Springer-Verlag,Berlin, Germany: NATO ASI Series 1990;
60:333-361.11. Jiang H, Robb PA, Holton KS. A new ap-
proach to 3-D registration of multimodalitymedical images by surface matching. SPIE1992; 1808:196-213.
12. Penfield W, Rasmussen T. The cerebralcortex of man. New York, NY: MacMillan,1950; 57.
13. Belliveau JW, Kennedy DN, McKinstry RC,
et at. Functional mapping of the human
visual cortex by magnetic resonance imag-ing. Science 1991; 254:716-719.
14. Fox PT, Raichle MR. Focal physiologicaluncoupling of cerebral blood flow and oxi-dative metabolism during somatosensorystimulation in human subjects. Proc NatIAcad Sri USA 1986; 83:i140-1144.
15. Brindle KM, Brown FF, Campbell ID, et al.Application of spin echo nuclear magneticresonance to whole-cell systems. Biochem1979; 180:37-44.
16. Thulborn KR, Waterton JC, Matthews PM,et al. Oxygenation dependence of thetransverse relaxation time of water protonsin whole blood at high field. Biochim Bio-phys Acta 1982; 714:265-270.
17. Ogawa S, Lee TM, Kay AR, et al. Brainmagnetic resonance imaging with contrastdependent on blood oxygenation. ProcNatI Acad Sci USA i990; 87:9868-9872.
18. Ogawa S, Lee TM, Nayak AS, et al. Oxy-genation-sensitive contrast in magneticresonance image of rodent brain at highmagnetic fields. Magn Reson Med 1990;14:68-78.
19. Turner R, Le Bihan D, Moonen CTW, Des-pres D, Frank J. Echo-planar time course
MRI of cat brain oxygenation changes.Magn Reson Med 1991; 22:159-166.
20. Ogawa 5, Menon RS, Tank DW, et at.
Functional brain mapping by blood oxy-genation level-dependent contrast mag-netic resonance imaging. Biophys J 1993;64:803-812.
21. Weisskoff RM, Boxenman JL, Zvo CS, RosenBR. Endogenous susceptibility contrast:principles of relationship between bloodoxygenation and MR signal change. Pne-sented at the Functional MRI of the BrainWorkshop, Arlington, Va,June 17-19, 1993.
22. LeBihan D,Jezzood P, Turner R, CuenodeC, Zffiro T. Analysis offunctional MRimages with z maps: work in progress.JMRJ 1993; 3(P):141.
23. Bandettini PA,Jesmanowicz A, Wong EC,Hyde JS. Processing strategies for time-course data sets in functional MRI of thebrain. Magn Reson Med 1993; 30:161-173.