fingertip representation in the human somatosensory cortex ... · primary somatosensory cortex...

Fingertip Representation in the Human Somatosensory Cortex:An fMRI Study

Patricia A. Gelnar,* Beth R. Krauss,* Nikolaus M. Szeverenyi,† and A. Vania Apkarian*,‡*Department of Neurosurgery, ‡Department of Computational Neuroscience, and †Department of Radiology,

SUNY Health Science Center, Syracuse, New York, 13210

Received August 19, 1997

Eight right-handed adult humans underwent func-tional magnetic resonance imaging (fMRI) of theirbrain while a vibratory stimulus was applied to anindividual digit tip (digit 1, 2, or 5) on the right hand.Multislice echoplanar imaging techniques were uti-lized during digit stimulation to investigate the organi-zation of the human primary somatosensory (SI) cor-tex, cortical regions located on the upper bank of theSylvian fissure (SII region), insula, and posterior pari-etal cortices. The t test and cluster size analyses wereperformed to produce cortical activation maps, whichexhibited significant regions of interest (ROIs) in allfour cortical regions investigated. The frequency ofsignificant ROIs was much higher in SI and the SIIregion than in the insula and posterior parietal region.Multiple digit representations were observed in theprimary somatosensory cortex, corresponding to thefour anatomic subdivisions of this cortex (areas 3a, 3b,1, and 2), suggesting that the organization of thehuman somatosensory cortex resembles that describedin other primates. Overall, there was no simple medialto lateral somatotopic representation in individualsubject activity maps. However, the spatial distancebetween digit 1 and digit 5 cortical representationswas the greatest in both SI and the SII region withinthe group. Statistical analyses of multiple activityparameters showed significant differences betweencortical regions and between digits, indicating thatvibrotactile activations of the cortex are dependent onboth the stimulated digit and cortical regioninvestigated. r 1998 Academic Press

INTRODUCTION

The first cortical maps of the human body representa-tion in the somatosensory cortex were published byPenfield and colleagues (Penfield and Boldrey, 1937;Penfield and Rasmussen, 1950). These homunculi wereproduced using electrical stimulation of the corticalsurface of the convexity of the postcentral gyrus inawake epilepsy patients. More recent intraoperative

mapping techniques have utilized cortical evoked poten-tials and resulted in improved results (e.g., Woolsey etal., 1979; Allison et al., 1989a,b, 1991). These intraop-erative procedures usually are performed under severetime constraints, using low impedance surface elec-trodes and, thus, these mappings have not yielded adetailed cortical map of human body representation. Asa result, Penfield and Rasmussen’s published homuncu-lus from 1950 is still the one most often used clinically.The primary somatosensory cortex (SI) of nonhumanprimates has been shown to be much more complexthan Penfield’s representations suggest (e.g., Mer-zenich et al., 1978; Kaas et al., 1979). These studiesreveal that each cytoarchitectonic subdivision of theprimary somatosensory cortex (areas 3a, 3b, 1, and 2)has its own fairly complete representation of the bodysurface.

New noninvasive imaging techniques offer the abilityto study the somatosensory cortex in the awake humanin more detail. Increased cortical activity in humans inresponse to cutaneous somatosensory or median nervestimuli has been reported using positron emissiontomography (PET) (e.g., Roland, 1981; Fox et al., 1987;Meyer et al., 1991; Seitz and Roland, 1992; Burton etal., 1993, 1997; Coghill et al., 1994), single photonemission computed tomography (SPECT) (Apkarian etal., 1992), magnetoencephalography (MEG) (Okada etal., 1984; Huttunen, 1986; Huttunen et al., 1987; Suk etal., 1991; Baumgartner et al., 1991; Forss et al., 1994;Gallen et al., 1993, 1994, 1995), somatosensory evokedpotentials (SEPs) (Allison et al., 1989a,b, 1991; Buch-ner et al., 1994, 1995), and functional magnetic reso-nance imaging (fMRI) (Hammeke et al., 1994; Puce,1995; Yetkin et al., 1995; Lin et al., 1996). To date thespatial resolution of PET and SPECT has not beensufficient to resolve individual digits in the somatosen-sory cortex of humans. The determination of corticalsomatotopy using PET has been made based upon theability to differentiate between different body regions,i.e., hand vs foot (Fox et al., 1987). Fox et al. (1987)showed distinct SI cortical regions activated with cuta-

NEUROIMAGE 7, 261–283 (1998)ARTICLE NO. NI980341

261 1053-8119/98 $25.00Copyright r 1998 by Academic Press

All rights of reproduction in any form reserved.

neous vibratory stimulation of the lip, hand, and foot ina mediolateral representation consistent with Penfieldmaps. In a subsequent reanalysis of this same data(Burton et al., 1993), the authors show activations inthe parietal operculum, assumed to be SII, and in theinsula. MEG and SEP have both reported successfuldifferentiation of the cortical representation of indi-vidual digits in the human primary somatosensorycortex (Suk et al., 1991; Baumgartner et al., 1991;Raichle, 1994; Buchner et al., 1995).

The temporal and spatial resolution of functionalmagnetic resonance imaging (fMRI) is an improvementupon that of PET or SPECT (Ogawa et al., 1990;Belliveau et al., 1991; Stehling et al., 1991), but theability of fMRI to differentiate between nearby bodyrepresentations has not been determined, e.g., betweenadjacent digits. The minimum spatial cortical separa-tion identified between fMRI activations for individualdigits delineates the functional spatial resolution of thecurrent technology, for the given imaging parametersused. Previous PET studies (Roland, 1981; Fox et al.,1987; Meyer et al., 1991; Burton et al., 1993) investigat-ing tactile stimulation revealed single foci of activity inSI, whereas a recent fMRI study (Lin et al.,1996)reported multiple foci, suggesting that the spatialresolution of fMRI may be adequate to differentiatebetween digits in SI. In the owl monkey the corticalsurface devoted to an individual digit within areas 3band 1 is on the order of 0.5 to 1 mm2 (Merzenich et al.,1978; Jenkins et al., 1990). Given the difference in brainsize between macaques and humans, the expectedcortical area dedicated to a single fingertip should beabout 5 mm2, which should be within the resolvingpower of fMRI with its higher spatial resolution. In anattempt to test the functional spatial resolution offMRI, we examined the ability of our imaging systemutilizing a high spatial resolution (1.56 3 1.56 mmpixel size) to study the representation of individualfingers within the human somatosensory cortices.

Different studies using fMRI have analyzed brainactivations with a variety of statistical approaches, e.g.,percentage change in signal intensity (Ogawa et al.,1992; Turner et al., 1993), correlation to the stimulus(correlation coefficient) (Rao et al., 1995; Lin et al.,1996; Flament et al., 1996), t values (Kim et al., 1994;Kami et al., 1995; Davis et al., 1995; Flament et al.,1996), z values (Ramsey et al., 1996), or more recentlystatistical parametric mapping (Sorenson and Wang,1996). Currently, there is no consensus as to the bestmeasure of fMRI activity. Therefore, a secondary pur-pose of this study was to investigate three activityparameters, t value, size in voxels, and their product (tvalue 3 size), in order to determine which can bestdifferentiate between fMRI activity changes.

METHODS

Eight normal right-handed volunteers participatedin this study. The general purpose and the procedureswere explained to the subjects. All subjects were atleast 18 years of age and signed a consent form. Allprocedures were approved by the Institutional ReviewBoard. Each individual underwent a single scanningsession which lasted approximately 1.5 h and consistedof a high-resolution anatomical scan in the sagittalplane, a high-resolution anatomical scan of the entirebrain in the coronal plane (6-mm slices, eight of whichcorresponded to those imaged in the functional scans),a flow-weighted scan for identification of cortical ves-sels, and three functional imaging series using EPIpulse sequences. Separate functional series were per-formed for each digit. In each functional imaging seriesa vibratory stimulus was applied to an individual digittip on the right hand, each lasting 7 min. During thefirst functional imaging series the thumb (D1) wasstimulated, the index finger (D2) during the secondfunctional series, and the fifth digit (D5) during thethird. The control for each series was rest. There was a14-min pause between each functional imaging series.

A new vibratory stimulator had to be designed forthis study due to the inability to introduce metallic orelectrical materials into the MRI suite. The stimulatorwas a plastic cylinder with an internal pneumaticallydriven wheel. The frequency and the amplitude of thestimulus were controlled by the applied air pressureand the mass of the eccentric lead weight attached tothe wheel, respectively. Both the air pressure andweight were maintained constant throughout the study.The stimulator delivered vibrations of mixed frequen-cies where the dominant frequency is 50 Hz with higherharmonics. The amplitude of the vibration was up to 2mm. The individual’s hand was positioned so that onlythe tip of the stimulated digit was touching the vibrat-ing surface. The manually controlled vibratory stimu-lus was on for 35 s and off for 35 s, for a total of sixcycles.

Scanning Sequence Procedure

All fMRI experiments were performed on a 1.5 TGeneral Electric (Signa) clinical imaging instrumentequipped with an Instascan resonant gradient acces-sory from Advanced NMR Systems Inc. thus allowingthe acquisition of both conventional and echoplanarimages. In order to improve signal-to-noise ratio asingle 5-in. circular surface coil was used.

During each imaging session the subject was posi-tioned on the scanner bed and the surface coil waspositioned over the parietal cortex, contralateral to thestimulated hand, and oriented parallel to the long axisof the magnet. The subject’s head and surface coil wereimmobilized using a vacuum bean bag (Olympic Vac-

262 GELNAR ET AL.

Pac; Olympic Medical) shaped to the individual’s head.Sagittal, high-resolution and flow-weighted images wereobtained with conventional pulse sequences while thefunctional images were obtained using echo planarpulse sequences. The high-resolution, flow-weighted,and functional scans were all performed at the sameslice locations. The scanning sequence proceeded asfollows:

(1) T1-weighted multislice spin echo scout images(TR 5 300 ms; TE 5 12 ms; 2 NEX; 256 3 256 matrix;FOV 5 20 3 20 cm) were obtained in the sagittal planeand prescription of subsequent coronal images wereperformed using the most midline sagittal slice (Fig. 1).The primary cortical areas of interest were the somato-sensory regions, so the middle third of the braincontralateral to the stimulation site was imaged. Eightslice locations were selected for subsequent EPI func-tional scans (each 6.0 mm thick with a 0.5-mm gapbetween slices). In each study the first slice was located6.5 mm posterior to the anterior commissure.

(2) High-resolution coronal multislice spin echo im-ages (TR 5 500 ms; TE 5 12 ms; 2 NEX; 256 3 256matrix; FOV 5 20 3 20 cm) of the entire brain wereobtained with 6.0-mm slice thickness and a 0.5-mm gapbetween slices. This set included the eight functionalslice locations, and was obtained for anatomical localiza-tion of the areas of functional activation.

(3) A flow-sensitive imaging sequence (SPGRASS;TR 5 33 ms; TE 5 6 ms; 4 NEX; Flip angle 5 60°;256 3 256 matrix; FOV 5 20 3 20 cm) was performedat the eight functional slice locations for localization ofthe large in plane cortical vessels.

(4) Functional imaging scans were then performedusing the following EPI GRE acquisition sequence:TR 5 3500 ms; TE 5 60 ms; flip angle 5 90°; NEX 5 1(NEX 5 2 for one subject); repetitions per slice 5 10 (5for the subject with 2 NEX); matrix 5 256 3 128;FOV 5 40 3 20 cm. This results in a voxel size of 1.56 31.56 3 6.00 mm. Ten images per slice location areacquired during each control and stimulus state and sixcycles of control and stimulus are performed during afunctional imaging series. Altogether 960 functionalimages were obtained at the eight slice locations in asingle functional imaging series. This results in 120images per slice location (60 control images and 60stimulus images) for use in the statistical analysis.

Data Analysis

The functional images were analyzed on a SUNSPARC-10 workstation utilizing software developed byour group.

Prior to statistical analysis, in-plane head movementwas corrected by reregistering all images at each slicelocation using a dot product maximization procedure.The first image at each slice location was chosen as thereference image for that slice location. Each subse-

quent image at that slice location was translated by onepixel in all directions (8) in the x and y, and eight dotproducts were calculated with the reference image. Themaximum dot product was then chosen. This procedurewas repeated until the untranslated configuration ofthe image produced the largest dot product. Thisalgorithm can correct exactly within plane shifts whenthese shifts occur at multiples of single pixels but is lessaccurate for shifts that are smaller than a single pixel.

To assure that the images were properly classified asbelonging to stimulus or control periods, images col-lected at the stimulus–control and control–stimulustransitions were discarded. As a result, of the total 960images collected in a functional imaging series, 96images were not used for further statistical analyses.

A region of cortex located underneath the 5-in. coilwas selected manually on the first image at each slicelocation. The mean intensity of this defined region wascalculated for all of the 120 images obtained at eachgiven slice location. Images with deviations larger thantwo standard deviations from the mean were attributedto artifacts, e.g., insufficient correction for head move-ment, homogeneity effects, and thus discarded as outli-ers. On average 15 (range 0–41) images were discardedper functional imaging series.

Further image analysis was performed on a pixel-by-pixel basis, and proceeded as follows:

1. Significant activity was based upon unpaired ttests performed on individual pixels. Multiple t testswere performed on each functional data series produc-ing stimulus t maps (stimulus vs control) and noise tmaps (control vs control, and stimulus vs stimulus) foreach slice location. In the t map image each pixel isassigned a value determined by the t value for thatpixel. A t value correlating with a P , 0.01 is defined asthe threshold for significant change.

2. Given the large number of pixels in each image,there is a sufficient probability that many pixels wouldpass the selected criterion level (P , 0.01) by chance.However, this probability decreases sharply when thenumber of contiguous pixels in a cluster examined isincreased. Since representation of function in the cor-tex is spatially continuous (nearby regions subservesimilar functions), functional cortical activity is morelikely represented by clusters of contiguous pixels. Anoise analysis was performed in order to determine theappropriate cutoff for number of contiguous pixels in acluster. Twenty separate noise t maps were generated.For the 10 control noise t maps, the control images wererandomly regrouped into two categories and comparedwith each other. The same procedure was used on thestimulus images to produce 10 stimulus noise t maps.The t value criterion for stimulus-control t maps wasapplied to these noise t maps. The regions of activitythat passed this criterion (ROIs) in the noise t mapswere used to generate a frequency distribution of the

263FINGERTIP REPRESENTATION IN HUMAN CORTEX

number of contiguous (touching) pixels (Fig. 2) similarto that performed by Roland et al. (1993). This distribu-tion represents the size (number of contiguous pixels)dependence of noise at the defined criterion, for eachfunctional imaging series. A second cutoff criterion fordefining significant activity is determined by the noisecluster distribution. The cluster sizes defining the top1% of the noise distribution were used as the cutoffvalue for cluster size.

3. Based on 1 and 2, significant activation wasdefined as clusters of contiguous pixels with significantt values (P , 0.01), and significantly large cluster sizes(99th percentile). The cluster cutoff value was usually 3pixels, except in one functional imaging series where itwas 5 pixels. The clusters that passed significance werethen superimposed on high-resolution images. Theseimages make up the cortical activity maps, and areused to identify the anatomical locations of the activa-tion clusters (significant ROIs). Further hypothesis-driven analysis (defined in 5, 6, and 7 below) wasperformed upon significant ROIs located within theprimary and secondary somatosensory cortices, theinsula, and the posterior parietal cortex (areas 5/7 and40).

4. Two investigators cooperatively identified func-tional cortical subdivisions based upon major sulci and

gyri, in multiple planes, and compared each slice withthe Tailaraich–Tournoux stereotactic atlas (Talairachand Tournoux, 1988). Significant ROIs located in theareas of interest were further selected by agreement ofthree investigators. Those ROIs obviously located inwhite matter or outside the brain were excluded.

5. An activated area (ROI) has multiple descriptors.Here we examined t value, size, and their product,defined as activation index (AI), in relation to the digitsstimulated and the brain regions activated. The mean tvalue, size, and activation index were determined forall significant ROIs located in the cortical regions ofinterest: SI, the SII region, insula, and posterior pari-etal cortex, in each functional series (D1, D2, and D5). Atwo-way repeated measures ANOVA (two-way RMANOVA) across subjects was then performed on themean values for size, t value, and activation indexusing cortical regions and digits as independent fac-tors.

6. The center-of-mass coordinates (CM) for all signifi-cant ROIs were calculated by averaging the x and ycoordinates of all of the pixels contained within eachindividual ROI. These CM coordinates were then repre-sented on the high-resolution images for each subjectand used for determining spatial separation betweendigit representations (see below). The CM coordinates

FIG. 1. Sagittal section of brain with lines depicting the center of the eight slice locations utilized in functional imaging studies. Theorange arrow indicates the anterior commissure.

264 GELNAR ET AL.

FIG

.3.

Act

ivat

ion

tmap

sin

one

subj

ect

due

tost

imu

lati

onof

the

tip

ofdi

git

1(D

1).S

lice

1(s

1)is

mos

tan

teri

oran

dsl

ice

8(s

8)is

mos

tpo

ster

ior.

Th

eor

ange

arro

ws

den

ote

the

sylv

ian

fiss

ure

and

the

gree

nar

row

sde

not

eth

ece

ntr

alsu

lcu

s.L

evel

sof

acti

vity

ofin

divi

dual

pixe

lsw

ith

inth

esi

gnifi

can

tR

OIs

are

repr

esen

ted

byco

lor

asin

dica

ted.

Act

ivit

ylo

cate

dw

ith

inar

eas

3b,1

,an

d2

are

labe

led

a,b,

and

c,re

spec

tive

ly.


for each subject were also approximated to stereotacticcoordinates (Talairach and Tournoux, 1988), based onthe identified landmarks on the high-resolution imagesin each slice.

7. Within each individual, the spatial distance (inmm) between the CM coordinates for each significantROI due to digit stimulation was calculated as follows:distance between individual digit 1 ROIs and digit 2ROIs (d1–2), distance between individual digit 1 ROI’sand digit 5 ROI’s (d1–5), and distance between indi-vidual digit 2 ROI’s and digit 5 ROI’s (d2–5). Thesedistances were calculated within subgroupings in eachcortical region based upon the individual subject’shigh-resolution activation maps.

Identification of Cortical Regions

The postcentral gyrus will be referred to as primarysomatosensory cortex (SI). The upper bank of thesylvian fissure in the parietal cortex (i.e., posterior tothe central sulcus) contains multiple body representa-tions including the second somatosensory cortex (SII),parietal ventral area (PV), and retroinsular area andpossibly others (Krubitzer and Kaas, 1990; Krubitzerand Calford, 1992; Krubitzer et al., 1995). For thepurposes of this study the upper bank of the sylvianfissure in the parietal cortex was defined as the SIIregion. The insular region was defined as the limen

insula, the cortex medial to the circular sulcus of Reil,and the cortex lateral to the superior limiting sulcus.The cortex lateral to the superior limiting sulcus wasincluded in the insular region based on recent anatomicstudies in the monkey (Craig et al., 1995). The posteriorparietal region was defined as parietal cortex posteriorto the postcentral sulcus, including areas 5, 7, and 40.

The primary somatosensory cortex of each subjectwas divided into subregions in a manner similar to thatperformed by Allison et al. (1989a), which was basedupon studies in humans and Old World monkeys (Pow-ell and Mountcastle, 1959; Vogt, 1928), and in referenceto the Talairach–Tournoux atlas (1988). Area 3a isdefined as the region at the depths of the central sulcus,and 3b is located in the posterior wall of the centralsulcus. Area 1 is located primarily in the anterior crownof the postcentral gyrus, and the posterior one-half ofthe crown of the postcentral gyrus and the anterior wallof the postcentral sulcus was defined as area 2.

It should be emphasized that the above parcellationsof the cortex are approximations based upon anatomi-cal landmarks. The borders of the SI subdivisionsoccasionally were guided by the activation patterns. Allcortical identifications were performed by the agree-ment of two investigators. The anatomic terminologyused follows standard nomenclature (for a useful guide,see Jackson and Duncan, 1996).

FIG. 2. Activation noise analysis in one subject for three consecutive scans where digit 1 (D1), digit 2 (D2), or digit 5 (D5) was stimulated.For each scan 10 control and 10 stimulus noise t maps were generated by randomly regrouping the corresponding images into two categories.The distribution of the number of clusters of a given pixel size passing the t value criterion (P , 0.01) is shown for these 20 noise t maps. Thecluster size defining the top 1% of this distribution corresponds to 3 pixels for all three scans (arrow).

266 GELNAR ET AL.

RESULTS

Comparison of Activation Maps Across Digitsfor Individual Subjects

A highly consistent pattern of activation was ob-served in individual subjects when digit activationmaps were compared. Cortical activity maps in atypical subject in response to D1, D2, and D5 stimula-tion are shown in Figs. 3–5. In this subject the majorityof activity was noted in SI (slices 3–8) and the SIIregion (slices 3–5).

SI

In general the same locations in SI were activeduring stimulation of all three digits, although the sizeof the activations and the exact positions varied. Mostof the SI activity is located in slices 4 and 5. In slice 4, (i)a region on the posterior bank of the central sulcus,area 3b (indicated as a in Figs. 3–5), is active for bothD1 and D2 but is absent for D5; (ii) a region on theconvexity of the postcentral gyrus, area 1 (indicated asb in Figs. 3–5), is active for both D2 and D5, while anarea slightly more medial to this is active for D1; and(iii) the same region in the posterior portion of thepostcentral gyrus, area 2 (indicated as c in Figs. 3–5), isactive for all 3 digits. In slice 5, the same 2 regions(medial region is area 1 (a) and more lateral is area 2(b)) on the postcentral gyrus are active for all 3 digits,but the area 2 activity secondary to D2 stimulation isslightly more medial than the others.

SII Region

The activity was located in slices 3–5. In slice 3 themidportion of the upper bank of the Sylvian fissure isactive with D2 and D5 stimulation, but the D5 activityis located slightly more lateral. In slices 4 and 5, the SIIregion activity was noted only with D5 stimulation.Activity located outside the cortex within the Sylvianfissure, in slice 4 with both D1 and D2 stimulation,were not counted.

Insula

Activity in the insula was noted at one slice locationonly (slice 2). The same insular region is active withboth D2 and D5 stimulation (Figs. 4 and 5).

Posterior Parietal

No posterior parietal activity was noted in thissubject.

Activity was also noted in the primary motor cortex(precentral gyrus in slices 4 and 5), this is likely due tothe fact that the subjects were required to hold theirfinger on the vibratory stimulator during the imagingsessions.

Unique characteristics were noted in the activationpatterns of some subjects. Certain individuals hadmore pronounced activity in the SII region compared toSI, and some had more insular and posterior parietalactivity than others. In two subjects the SII regionactivation was more than SI for all three digits, asmeasured by size, t values, and activation index. Twosubjects had more pronounced insular activity. One ofthese subjects had consistent activation of multipleinsular areas, while the other individual had the samearea active in response to stimulation of two digits andanother insular region active in response to stimulationof the third digit. Two subjects had no posterior parietalactivity at all and one subject had minimal posteriorparietal activity in response to stimulation of one digit.

Time Course of Activations

The time course of cortical activations is shown fortwo subjects in Fig. 6. The left panels are for the subjectwhose t maps are presented in Figs. 3–5. The rightpanels are from a subject where all four cortical areasstudied were activated. These time curves indicate thatthe cortical activity reaches peak activation within twoto three time steps from the start of the stimulus,undergoes a slight adaptation, and then maintains theincreased activity for the duration of the stimulus. Inthe control phase, the activity decreases at a relativelyconstant rate and reaches baseline only in the last fewcontrol images. The peak-to-peak percentage change insignal intensity ranges from 2 to 8%. However, thelarge percentage changes do not necessarily correspondto large t values. For example, in the left bottom panel,digit 5 stimulation (Fig. 6), the largest percentagechange was in the ROI in the SII region (about 5%), itscorresponding t value was 6.9. The smallest percentagechange in this panel was in SI (about 2%), its corre-sponding t value was 11.2.

The left top panel indicates activations in two ROIslocated in SI in consecutive slices. The two curves peakwith a 7-s difference in delay. If this difference is takenas reflecting the variability of the response peak, thenthe activation time courses of the different corticalareas do not differ from each other in their times toreach peak activations.

Incidence of Cortical Activations

The cortical activation maps to vibratory stimulationof D1, D2, and D5 exhibited significant ROIs in all fourcortical regions investigated: SI, the SII region, insula,and posterior parietal cortices and is summarized inTable 1. Significant activity due to stimulation of atleast one digit was observed in all subjects in SI, the SIIregion, and the insula, but only in 6 of the 8 subjects inthe posterior parietal area.

Table 1 shows that D1, D2, and D5 were represented


FIG

.4.

Act

ivat

ion

tmap

sin

the

sam

esu

bjec

tas

inF

ig.3

due

tost

imu

lati

onof

the

tip

ofdi

git

2(D

2).S

eeF

ig.2

for

deta

ils.

268 GELNAR ET AL.

FIG

.5.

Act

ivat

ion

tmap

sin

the

sam

esu

bjec

tas

inF

ig.3

due

tost

imu

lati

onof

the

tip

ofdi

git

5(D

5).S

eeF

ig.3

for

deta

ils.


FIG. 6. Time course of cortical activations in two subjects. The left column shows cortical activation in one subject, corresponding to the tmaps shown in Figs. 3–5. The percentage change shown for stimulating digit 1 (D1; top panel) indicates activation time course for 2 ROIs in SI.The slice 4 SI activation corresponds to the ROI designated as c in Fig. 3, s4; it is made up of 16 pixels and has a t value of 6.3. The slice 5 SIactivation corresponds to the ROI designated as b in Fig. 3, s5; it is made up of 10 pixels and has a t value of 6.3. The percentage change forstimulating digit 2 (D2; middle panel) indicates activation time courses for ROIs in SI, insula (INS), and the SII region (SII). The SI ROI had at value of 4.2; the insula had a t value of 6.0; and the SII had a t value of 5.3. Bottom panel shows activation time courses for the same corticalregions when digit 5 (D5) was stimulated. The t values in this case were 6.9 for SII, 5.0 for insula, and 11.2 for SI. The right column showsactivation time courses in a second subject where all four cortical areas studied were activated. Top panel shows percentage activation whendigit 1 (D1) was stimulated for ROIs in insula, the SII region, SI, and the posterior parietal cortex (PP). Lower panel shows activation whendigit 5 was stimulated for an ROI in the SII region comprising of 39 pixels with a t value of 13.3. In this case the activation is presented in MRsignal intensity across all six control and stimulus cycles (120 images). The percentage change activations are the mean changes when all sixcontrol and stimulus cycles are averaged, and the MR signal in the ninth control image is set to 100%. The range of the standard error of themean percentage change was 0.2–0.8. Thick lines on the x axes indicate stimulus periods.

270 GELNAR ET AL.

more often in SI and the SII region than in insula orposterior parietal cortices. All three digits were repre-sented in the primary somatosensory cortex in all 8subjects (8/8) and in the SII region in 6 of 8 subjects.The insular cortex and posterior parietal area had all 3digits represented in only 2/8 subjects. A one-waynonparametric (Kruskal–Wallis) ANOVA of the occur-rence of significant activity in each cortical region overall 3 digits showed highly significant differences be-tween cortical regions (H 5 17.0, P , 0.0007). Post-hocpairwise multiple comparison (Student–Newman–Keuls method) showed statistically significant differ-ences in incidence of activity between SI versus insulaand posterior parietal, and the SII region versus insulaand posterior parietal. Thus, the frequency of signifi-cant activity due to stimulation of the digits was muchhigher in SI and the SII region than in the insula andposterior parietal region.

Multiple Activation Foci within SI

The vibratory cortical activation maps revealed mul-tiple foci of activation within the primary somatosen-sory cortex. In an effort to determine whether thesemultiple foci of activity were homologous to the SIsubregions identified in nonhuman primates, each ROIwas assigned to a single SI subregion (3a, 3b, 1, or 2).The vibratory cortical activation maps for all subjectsexhibited ROIs in all SI subregions investigated (Table2). There were no major differences in individual digitrepresentation (D1, D2, or D5) frequency within the SIsubregions. However, there were significant differencesin incidence of activity across subregions (3a, 3b, 1, and2). A one-way nonparametric (Kruskal–Wallis) ANOVAof the occurrence of significant activity in each subre-gion of SI over all three digits showed highly significantdifferences between subregions (H 5 40.1, P , 0.0001),with a post-hoc pairwise multiple comparison (Student–Newman–Keuls method) showing statistically signifi-cant differences in incidence of activity between area 1versus areas 3a and 3b, and area 2 versus areas 3a and3b. Thus, the frequency of significant activity due tostimulation of the digits was much larger in areas 1 and2 than in areas 3a and 3b.

Somatotopy

Figure 7 is an example of the center-of-mass coordi-nates for D1, D2, and D5 for one subject. In this subject,D1 and D5 are represented in the insula (s1 and s4), allthree digits are found in the SII region (s2–s5), multiplerepresentations of all three digits are found in SI, andonly digits 1 and 5 in the posterior parietal cortex.Overall there was no simple medial-to-lateral somato-topic representation in individual subject center-of-mass maps, as illustrated in Fig. 7. However, multiplegroupings of digit representations did appear to existwithin SI and SII region.

The center-of-mass coordinates for each significantcluster in all subjects are represented upon the appro-priate coronal talairach atlas section in Fig. 8. Again,no clear medial-to-lateral somatotopic representationwas observed. Activity in SI is concentrated in theregion from 16 to 28 mm posterior to AC. All three digitshave ROIs throughout this region. However, the high-est incidence of D5 representation is in the mostposterior slice (228 mm from AC); the highest inci-dence for D1 is around 224 mm from AC; and the D2representation seems to be more equally distributedthroughout the region. This suggests that D5 is locatedmore medially than D1 or D2 in SI. Activity in the SIIregion extends from 8 to 28 mm posterior to AC. In theSII region, the incidence of representation for all threedigits increases more posteriorly. Activity in the insularregion extends from 4 to 20 mm posterior to AC, but isconcentrated in the region from 4 to 8 mm posteriorfrom AC. All three digits seem equally represented inthis region. Activity in the posterior parietal cortexspreads from 28 to 50 mm posterior to AC and isdominated by D1 and D5 activity.

Since no simple medial-to-lateral somatotopic repre-sentation was seen when examining the individualcenter-of-mass maps and the Talairach maps, the spa-tial distance between significant clusters (d1–2, d1–5, andd2–5) within the subregions of SI and within the othercortical regions were compared to determine whether amore subtle somatotopic representation was present.

The spatial distance differences between the center-of-mass coordinates for the activated ROIs for indi-vidual digits in SI and the SII region are shown in Table3. The distance between D1 and D5 ROIs (d1–5) was the

TABLE 1

The Presence of Significant Activity in Four Cortical Regionsfor the Three Digits Stimulated in Eight Subjects

SI SII region InsulaPosteriorparietal

D1 8 7 4 5D2 8 7 4 3D5 8 7 8 5All 3 digits represented 8 6 2 2

TABLE 2

The Presence of Significant Activity in Four Subregions of SIfor the Three Digits Stimulated in Eight Subjects

3a 3b 1 2

D1 3 2 6 6D2 0 1 6 7D5 1 2 7 8All 3 digits represented 0 0 4 6


FIG

.7.

Th

ece

nte

rof

mas

slo

cati

ons

for

sign

ifica

nt

RO

Isin

one

subj

ect

due

tost

imu

lati

onof

the

tips

ofdi

gits

1,2,

and

5as

indi

cate

d.F

orde

tail

sse

eF

ig.3

.

272 GELNAR ET AL.

FIG

.8.

Th

ece

nte

r-of

-mas

sco

ordi

nat

esre

pres

ente

din

the

Tala

raic

hat

las

for

each

sign

ifica

nt

clu

ster

inal

lfu

nct

ion

alim

agin

gse

ries

inal

lsu

bjec

ts.T

he

ante

rior

–pos

teri

orlo

cati

onof

the

sect

ion

inre

fere

nce

toth

eA

C(i

nm

illi

met

ers)

isin

dica

ted

inth

eu

pper

righ

t-h

and

corn

er(p

osit

ive

isan

teri

orto

the

AC

and

neg

ativ

eis

post

erio

rto

the

AC

).B

rodm

ann

sar

eas

are

indi

cate

don

the

cort

ex.T

he

oran

gear

row

sde

not

eth

esy

lvia

nfi

ssu

rean

dth

egr

een

arro

ws

the

cen

tral

sulc

us.

Dig

it1

isre

pres

ente

dby

gree

npi

xels

,dig

it2

isre

pres

ente

dby

yell

owpi

xels

,an

ddi

git5

isre

pres

ente

dby

red

pixe

ls.


greatest in both SI and the SII region. The spatialseparations between ROIs for individual digits showeda significant difference in SI between d1–5 and d2–5 (ttest, P , 0.02). No significant differences were found inthe SII region.

Statistical Analysis Using T Value as Dependent Variable

The mean t value across subjects was largest for D5and smallest for D2 (Fig. 9A). When comparing be-tween cortical regions the largest mean t value wasseen in SI, while the posterior parietal area had thesmallest t value (Fig. 9B). A two-way repeated mea-sures ANOVA (two-way RM ANOVA) of t values usingstimulated digits and brain regions as independentfactors showed a highly significant difference betweencortical regions (F 5 6.9, P , 0.002), but no significantdifference between digits (F 5 2.2, P . 0.15). Post-hocanalysis with the paired Bonferroni t test indicatedsignificant differences (P , 0.05) in t values between SIand posterior parietal cortex and between the SIIregion and posterior parietal cortex.

Statistical Analysis Using Size as Dependent Variable

The mean size (in pixels) across subjects was largestfor D5 and smallest for D2 (Fig. 9C). When comparingbetween cortical regions the largest mean size was seenin SI and the SII region, while the insula had thesmallest mean size (Fig. 9D). A two-way RM ANOVA ofsize using stimulated digits and brain regions as inde-pendent factors showed near significant differencesbetween both digits (F 5 3.2, P , 0.07) and corticalregions (F 5 2.6, P , 0.08).

We compared the total number of pixels activated ina functional series to the subject’s brain size (in cm2,biparietal distance 3 anterior–posterior distance).There was no discernible correlation between brain sizeand number of active pixels across subjects.

Statistical Analysis Using Activation Index as DependentVariable

The mean activation index across subjects was larg-est for D5 and smallest for D2 (Fig. 9E). When compar-

ing between cortical regions the largest mean activa-tion index was seen in the SII region, while the insulahad the smallest mean activation index (Fig. 9F). Atwo-way RM ANOVA of activation index with stimu-lated digits and brain regions as independent factorsshowed significant differences between digits (F 5 4.1,P , 0.04) but not cortical regions (F 5 1.4, P . 0.27).Post-hoc analysis using the paired Bonferroni t testrevealed no statistically significant differences.

Time-Dependent Changes in Cortical Activations

The functional images in one subject were subdividedinto two groupings based on their temporal relation-ship to the onset of the stimulus. Functional imagesfrom the first half of the six control and stimulusperiods (0 to 18 s of each cycle) were grouped together,constituting the early phase, and the images from thesecond half (18–35 s of each cycle) were grouped,constituting the late phase. Figures 10 and 11 show theearly phase (Fig. 10) and late phase (Fig. 11) activationmaps for D2 stimulation. The locations of activity inthese maps are similar to those described above for thetotal stimulus control durations. The active ROIs inboth the early and late phases were in correspondinglocations; however, both the size and t values of theactive ROIs were larger in the late phase. A similarincrease in cortical activity in the later phase was alsoseen for the other two digit representations.

DISCUSSION

A cutaneous low-threshold stimulus resulted in fMRIactivity in SI, the SII region, mid- and posterior insula,and posterior parietal cortex (areas 5/7 and 40). Both SIand the SII region were more consistently active thanthe insula and the posterior parietal cortex. Within SImultiple foci of activation were found corresponding toits four cytoarchitectural subdivisions, 3a, 3b, 1, and 2;most activity was concentrated in areas 1 and 2. Thetopography of digit representation showed that D5 waslocated most medially on the postcentral gyrus, and themean spatial separation was largest between D1 andD5. This finding is consistent with the medial-to-lateralcortical distribution noted in neuromagnetic and so-matosensory evoked potential (SEPs) studies (Okada etal., 1984; Huttunen et al., 1987; Baumgartner et al.,1991; Buchner et al., 1995). Of the three digits stimu-lated, D5 stimulation resulted in a higher number ofactivated pixels. Three activation parameters weretested for differentiating between brain regions anddigit stimulation. The t value differentiated betweencortical regions but not between digits, whereas activa-

TABLE 3

Mean Spatial Separation between Digit Representations inSI and SII Region (in mm) across All Eight Subjects

SI SII region

d1–2 8.67 (1.23) 5.66 (1.30)d1–5 9.52 (1.23) 9.90 (2.58)d2–5 6.14 (1.38) 6.78 (1.11)

Note. SEM in parentheses.

274 GELNAR ET AL.

tion index distinguished between digits but not corticalregions, and size did not differentiate either parameter.

Activity in SI

Increased activity in response to cutaneous stimula-tion has been demonstrated in SI in many functionalimaging studies such as: PET (Burton et al., 1997;Coghill et al., 1994; Seitz and Roland, 1992; Meyer et

al., 1991; Fox et al., 1987; Drevets et al., 1995; Roland,1981), MEG (Okada et al., 1984; Huttunen, 1986;Huttunen et al., 1987; Suk et al., 1991; Baumgartner etal., 1991; Forss et al., 1994; Gallen et al., 1993, 1994,1995), SEP (Allison et al., 1989a,b, 1991; Buchner et al.,1994, 1995), and fMRI (Hammeke et al., 1994; Puce,1995; Yetkin et al., 1995; Lin et al., 1996). The locationof activity in SI relative to the AC in this study wassimilar to that described by the PET studies. In our

FIG. 9. The distribution of three parameters used for statistical comparison of activity as a function of the two independent parameters,digits (A, C, and E) and cortical areas (B, D, and F). A and B show the mean and standard deviation (SD) of significant ROIs when t value isused as the dependent parameter, while C and D use the number of activated pixels, and E and F use the activation index as the dependentparameter.


FIG

.10.

Act

ivat

ion

tm

aps

inon

esu

bjec

tdu

eto

ate

mpo

ral

anal

ysis

ofco

rtic

alac

tivi

tyw

hen

digi

t2

(D2)

was

stim

ula

ted.

Act

ivat

ion

tm

aps

repr

esen

tth

eea

rly

phas

eof

cort

ical

acti

vity

,i.e

.,ac

tivi

tydu

eto

the

firs

th

alf(

0–18

s)of

the

con

trol

and

stim

ulu

spe

riod

s.S

eeF

ig.3

for

deta

ils.

276 GELNAR ET AL.

FIG

.11.

Act

ivat

ion

tm

aps

inth

esa

me

subj

ect

asin

Fig

.10

due

toa

tem

pora

lan

alys

isof

cort

ical

acti

vity

wh

endi

git

2(D

2)w

asst

imu

late

d.A

ctiv

atio

ntm

aps

repr

esen

tth

ela

teph

ase

ofco

rtic

alac

tivi

ty,i

.e.,

acti

vity

due

toth

ela

sth

alf(

18–3

5s)

ofth

eco

ntr

olan

dst

imu

lus

peri

ods.

See

Fig

.3fo

rde

tail

s.


study, SI digit region was mainly 216 to 228 mm fromAC compared to 27 in Meyer et al. (1991), 223 in Fox etal. (1987), 225 in Drevets et al. (1995), and 231 inCoghill et al. (1994).

Multiple foci of activity were consistently noted inthe primary somatosensory cortex in this study. Thelocation of these foci on the postcentral gyrus revealedfour regions, which are homologous to the known SIsubregions in monkeys (Merzenich et al., 1978). Otherinvestigators have also noted multiple foci of activity inSI cortex utilizing fMRI (Lin et al., 1996), PET (Burtonet al., 1997), and SEPs (Allison et al., 1989a, 1991).Both Lin et al. (1996) and Burton et al. (1997) noted twofoci of activity in SI, whereas we demonstrate fourseparate areas of activity in SI. In the fMRI study byLin et al. (1996) two foci of activity noted that the morerostral focus was in the depths of the central sulcus(3a). The second more caudal focus ‘‘appeared laterallyat the lips of this sulcus and extended posteriorly intothe postcentral sulcus,’’ implying activity in areas 3b, 1,and 2. So the same regions of SI were noted to be activein both studies, but in our study the activity was morediscrete. This difference is most likely due to the factthat multiple digits were stimulated in Lin et al. (1996)while only one digit tip was stimulated in this study.The results of Burton et al. (1997) showed results verysimilar to those of Lin et al. (1996); however, Burton etal. (1997) conducted a PET study with a resolution ofapproximately 11 mm FWHM, which makes it verydifficult to detect individual foci of activity across thepostcentral gyrus. Burton et al. (1997) proposed thatthese two foci of activity were consistent with theknown somatotopic representation in macaques. Ourresults suggest that the somatotopic representationwithin the SI subregions may be different than that innonhuman primates, similar to the differences seenbetween Old World and New World monkeys (Sur et al.,1982).

Most of the activity within SI was noted to be locatedin areas 1 and 2 in this study. This observation issurprising and must be due to the unique spatiotempo-ral characteristics of the stimulus used. Since thevibrotactile stimulus has most of the energy in the50-Hz range, it should preferentially activate rapidlyadapting type fibers in the skin. However, this stimula-tor has energy components at many higher frequenciesso it should also activate Pacinian-type receptors. Giventhat the stimulus has no discernible tactile direction orpattern components, and area 3b contains the simplesttype somatosensory neurons (Sur, 1980; Iwamura et al.,1983; Mountcastle, 1984; Merzenich et al., 1978) withno direction or orientation sensitivity (Costanzo andGardner, 1980; Whitsel et al., 1978, 1980; Iwamura andTanaka, 1978; Hyvarinen and Poranen, 1978), it wasexpected that areas 3b and 1 would be preferentiallyactivated. There are, however, adaptation differences

between neurons in area 3b and 1, as well as differencesin the incidence of slowly adapting, rapidly adapting,and Pacinian-type neurons (Phillips et al., 1988; Paul etal., 1972; Sur, 1980; Mountcastle, 1984; Prud’homme etal., 1994). Moreover, it was recently shown that cortical(Nicolelis et al., 1995; Lee and Whitsel, 1992) andsubcortical (Nicolelis et al., 1993) somatosensory re-sponses have complex spatiotemporal dynamics whichare highly dependent on the stimulus parameters.Therefore, the specific properties of the stimulus usedthat resulted in the given activation pattern remainunknown, and the more general issue of somatosensorycortical activity pattern dependence on stimulus param-eters remains to be studied.

Digit 5 was noted to have the largest activity asmeasured by all three activation parameters. Sincedigit 5 was always the last digit to be studied, weexamined the scanner properties as a function of timeto rule out scan ordering artifacts. The activation noiseanalysis and the brain versus background signal-to-noise ratios were mostly constant and independent ofscan order. Therefore, scanner peculiarities cannotaccount for the cortical activation dependence on thestimulated digit. Digit 5 having the largest activity wasan unexpected finding given the fact that area ofcortical representation in the somatosensory cortex ingeneral correlates with the number of receptors on thebody surface represented, and D5 should have fewerreceptors than D1 or D2 (Johannsson and Vallbo, 1979).In Penfield’s (1950) homunculus of SI, the largest areaof cortical representation was given to digit 1 and thesmallest digit cortical representation was associatedwith digit 5. Merzenich et al. (1987) have shown that inthe squirrel monkey the relative areas occupied by digit1 versus digit 5 is different between 3b and 1. In area 3bD5 had the smallest cortical area of representation,while in area 1 (in 4 of 5 squirrel monkeys) D5representation was greater than that for D1. Thus ourresults suggest that the area 1 digit representationmap seems to correspond to that seen in area 1 of thesquirrel monkey. A possible explanation for differencesin the size of activations for different digits could beexperience or preferential use of D1 and D2 over D5.However, the effects of experience in area 1, over thetime scale of a lifetime, remain unknown.

Asimple approach to unraveling the temporal dynam-ics of the cortical response is a comparison of the earlyversus late phase activations (within the same scan).This comparison was carried out in one subject andrevealed that the cortical activity has a strong timedependence, implying either temporal recruitment ofsurrounding neurons or reflecting the hemodynamicproperties of the response. At this time resolution therewas no evidence for shifts in cortical areas activated.

Analysis of distances between the significant ROIs inSI in this study reveals that the distances between

278 GELNAR ET AL.

representations of D1 and D5 were always the largest,while the distances between D2 and D5 were alwaysthe smallest. Approximation of the ROIs to Talaraichspace also revealed that the highest incidence of D5representation was located in the most posterior por-tion of the SI cortex, which would place it medial to bothD1 and D2. An MEG study by Suk et al. (1991) alsoutilizing vibratory stimulation of individual digits (D1,D2, and D5) in humans published very similar findings.Statistical analysis showed that the most significant(P , 0.05) difference in dipole source location wasbetween D1 and D5, with the former being superior tothe sources of the other two digits in all cases. Thesefindings are consistent with the crude homunculus ofthe human cortex published by Penfield (1950) andwith the results reported in other MEG and SEPstudies (Okada et al., 1984; Huttunen et al., 1987;Baumgartner et al., 1991; Buchner et al., 1995). Whenthe data for individual subjects were combined, nosimple medial to lateral representation was noted. Thecomplexity of the combined data was most likely due toindividual anatomical and functional variability, aswell as the presence of multiple active foci, all of whichobscured a simple medial to lateral somatotopic repre-sentation in SI. Buchner et al. (1995) measured SEPs inresponse to stimulation of individual digits (D1, D3,and D5) in eight normal human subjects. They foundthat in six of eight subjects the dipole source for thedigits was as follows from medial to lateral: D5 . D3 .D1, while in two subjects the mediolateral representa-tion was different. These results suggest that humanfingertip representation in SI may be more variablethan that seen in other primates.

Activity in the SII Region

Several PET (Seitz and Roland, 1992; Coghill et al.,1994; Burton et al., 1993, 1997) and MEG (Forss et al.,1994) studies have demonstrated activity in the upperbank of the Sylvian fissure in response to vibrotactilestimulation. Neurophysiologic studies have demon-strated that the upper bank of the Sylvian fissure inparietal cortex is a complex topographically organizedregion including several representations, the most well-known of which are the two mirror-reversed body mapsof SII and PV (Krubitzer and Kaas, 1990; Krubitzer andCalford, 1992; Krubitzer et al., 1995). Our inability toconsistently demonstrate multiple foci of activity inthis region is most likely due to the location and size ofthe SII region (upper bank of the Sylvian fissure is farfrom the surface coil), in combination with the mirror-reversed body maps. The location of activity in the SIIregion relative to the AC in this study spanned theregions described by the previous PET studies. In ourstudy the SII region activity in response to digitstimulation was quite extensive, extending from 28 to228 mm relative (posterior) to AC compared to the PET

studies with anterior–posterior coordinates for theparietal operculum activation: 24 in Burton et al.(1993), 221 and 226 in Coghill et al. (1994), and 223and 232 in Burton et al. (1997). The fMRI activityreported in this study involved all of the areas reportedto be active in the three PET studies. The large extentof activity in the SII region noted in this study and theprior PET studies is most likely due to the multiplerepresentations that exist on the upper bank of theSylvian fissure.

Activity in the Insula

Neuroanatomic studies reveal numerous somatosen-sory projections to the posterior one-third of the insula(Burton and Jones, 1976; Mesulam and Mufson, 1982;Mufson and Mesulam, 1982, 1984; Friedman et al.,1986;Krubitzer and Kaas, 1990; Krubitzer and Calford,1992; Apkarian and Shi, 1997). Electrophysiologic stud-ies have also described responses to somatic stimula-tion in and around the insular cortex (Robinson andBurton, 1980; Krubitzer et al., 1995). Both Robinsonand Burton (1980) and Krubitzer et al. (1995) demon-strated neuronal responses to tactile stimulation in theposterior one-third of the insula; in addition, Krubitzeret al. (1995) also noted physiologic responses to somaticstimulation in the insular fundus and lower bank of thelateral sulcus. The neurons in the peri-insular region ingeneral had larger receptive fields than those in SII andPV and a distinct representation of body parts (Kru-bitzer et al., 1995). PET studies using a stimulussimilar to the one utilized in this study reportedactivity in the insula (Coghill et al.,1994; Burton et al.,1993). Burton et al. (1993) demonstrated increasedrCBF localized primarily to the middle and posteriorportions of the insula. In our study only the posteriorone-half of the insula was imaged and digit representa-tion extended from 24 to 220 mm (posterior) from AC.In comparison Burton et al. indicate insular activity at216 and Coghill et al. indicate activity at 213 mmrelative to AC. In the present study only two individu-als had representation of all three digits in the insularcortex, the distance between digit representations inthe insula was less than in any other cortical region,and a somatotopic organization for the digits was notdetected in the insula.

Activity in Posterior Parietal Cortex

The posterior parietal region has been labeled thesensory association cortex due to bilateral afferentinput from somatosensory and visual cortices. Area 5neurons were first shown to exhibit complex multijointand multilimb receptive fields by Duffy and Burchfiel(1971), suggesting that an integration of the somatosen-sory information from SI and the SII region did exist.Cortical activity in the posterior parietal region has


also been noted previously in PET studies (Roland,1981; Seitz et al., 1991; Roland et al., 1989; Pardo et al.,1991). In these studies areas 5, 7, and 40 have beenassociated with simple and complex somatosensorystimulation, limb movement, and limb exploration.Due to the large number of SI and SII afferents to theposterior parietal cortex it is not surprising that asignificant amount of fMRI activity was noted in thisstudy. Areas 5, 7, and 40 all exhibited activations acrosssubjects but in general the posterior parietal activity asdemonstrated by the three analytic techniques (t value,size, and activation index) was always less than thatseen in SI and the SII region, was noted in fewersubjects, and had no somatotopic organization. Thelower levels of activity in this region may be primarilydue to the simple, uninteresting vibrotactile stimulus.Some of the activity in areas 5/7 may be also due to themotor requirements of the task.

Parameter Analysis

Three different parameters were used as a means forhypothesis testing. Each has its own merits. The t valuewas used instead of the percentage change in signalintensity because the former is a normalized unitlessmeasure related to the magnitude of the activationexpressed as a signal-to-noise ratio. Therefore the tvalue reflects the magnitude of change as well as ameasure of confidence regarding this change, and, asshown in Fig. 6, large percentage changes do notnecessarily correspond to large t values. It should benoted that the t value and correlation coefficient aredirectly related and, thus, interchangeable parameters.Another advantage of the t value is that the variabilityacross brain regions and subjects is compensated for bythe normalization with the standard deviation. This isespecially important when the scans are performedusing surface coils, which introduces a systematicgradient of variation in the signal-to-noise ratio as afunction of ROI distance from the coil. The secondparameter that was used here was size of the activeROI, which is a measure of the area of cortex activated.The third measure used was the activation index,which multiplicatively combines the magnitude (t value)and area (size in voxels) of activation, resulting in anintegrated activity parameter. The advantage of thethird parameter, not illustrated in this study (seeApkarian et al., 1998, in preparation), is that it can beused as a measure with which different conditions maybe compared for the whole brain, or for a specificportion of the brain, within and across subjects. Whenthese three parameters were used to compare brainactivity between stimulated digits and cortical regions,each indicated different statistical dependencies.

The results indicate significant differences in t valuesacross cortical regions and in activation index acrossdigits. The former suggests that when comparisons of

activations between cortical regions are performed, thet value may be the analytic parameter of choice, and thelatter suggests that comparison within a cortical regionmay be best performed using the activation index. Bothof these ideas remain to be validated in other tasks.

CONCLUSIONS

This functional MRI study of the human somatosen-sory cortex shows that the organization of the regionresembles that described in squirrel, owl, and macaquemonkeys. Multiple digit representations were observedin the primary somatosensory cortex, corresponding tothe four anatomic subdivisions of this cortex. Due toacross-subject variability and the spatial resolution offMRI, the mediolateral somatotopy of digit representa-tion could be detected only probabilistically within theprimary somatosensory cortex. Statistical comparisonof multiple activity parameters indicated that vibrotac-tile activations of the cortex are dependent on thestimulated digit and cortical region. Data were pre-sented suggesting that the cortical activations are notstatic and systematically change over the duration ofthe stimulus.

ACKNOWLEDGMENTS

The authors thank Michael Fonte, Sean Huckins and GwenTillapaugh-Fay for technical assistance. The project was funded inpart by the Department of Neurosurgery and a grant from NINDS(RO1 NS35115-02).

REFERENCES

Allison, T., McCarthy, G., Wood, C. C., Darcey, T. M., Spencer, D. D.,and Williamson, P. D. (1989a). Human cortical potentials evoked bystimulation of the median nerve. I. Cytoarchitectonic areas gener-ating short-latency activity. J. Neurophysiol. 62:694–710.

Allison, T., McCarthy, G., Wood, C. C., Williamson, P. D., and Spencer,D. D. (1989b). Human cortical potentials evoked by stimulation ofthe median nerve. II. Cytoarchitectonic areas generating long-latency activity. J. Neurophysiol. 62:711–722.

Allison, T., Wood, C. C., McCarthy, G., and Spencer, D. D. (1991).Cortical somatosensory evoked potentials. II. Effects of excision ofsomatosensory or motory cortex in humans and monkeys. J.Neurophysiol. 66:64–82.

Apkarian, A. V., Krauss, B., Szeverenyi, N. M., and Thomas, S.(1998). Sympathetically mediated chronic pain is a hyperfrontalstate. Manuscript in preparation.

Apkarian, A. V., and Shi, T. (1997). Thalamocortical connections ofthe cingulate and insula in relation to nociceptive inputs to thecortex. In Pain Mechanisms and Management (S. Ayrapetian, A. V.Apkarian, Eds.), IOS Press, Amsterdam.

Apkarian, A. V., Stea, R. A., Manglos, S. H., Szeverenyi, N. M., King,R. B., and Thomas, F. D. (1992). Persistent pain inhibits contralat-eral somatosensory cortical activity in humans. Neurosci. Lett.140:141–147.

280 GELNAR ET AL.

Baumgartner, C., Doppelbauer, A., Deecke, L., Barth, D. S., Zeitl-hofer, J., Lindinger, G., and Sutherling, W. W. (1991). Neuromag-netic investigation of somatotopy of human hand somatosensorycortex. Exp. Brain Res. 87:641–648.

Belliveau, J. W., Kennedy, D. N., and Mckinstry, R. C. (1991).Functional mapping of human visual cortex by magnetic resonanceimaging. Science 254:716–719.

Buchner, H., Adams, L., Knepper, A., Ruger, R., LaBorde, G., Gils-bach, G., Ludwig, I., Reul, J., and Scherg, M. (1994). Preoperativelocalization of the central sulcus by dipole source analysis of earlysomatosensory evoked potentials and three-dimensional magneticresonance imaging. J. Neurosurg. 80:849–856.

Buchner, H., Adams, L., Miller, A., Ludwig, I., Knepper, A., Thron, A.,Niemann, K., and Scherg, M. (1995). Somatotopy of human handsomatosensory cortex revealed by dipole source analysis of earlysomatosensory evoked potentials and 3D-NMR tomography. Elec-troenceph. Clin. Neurophysiol. 96:121–134.

Burton, H., and Jones, E. G. (1976). The posterior thalamic regionand its cortical projection in New and Old World monkeys. J. Comp.Neurol. 168:249–299.

Burton, H., McLeod, A. M., Videen, T. D., and Raichle, M. E. (1997).Multiple foci in parietal and frontal cortex activated by rubbingembossed grating patterns across finger pads: A positron emissiontomography study in humans. Cerebral Cortex 7:3–17.

Burton, H., Videen, T. O., and Raichle, M. E. (1993). Tactile-vibration-activated foci in insular and parietal opercular cortex studied withpositron emission tomography: mapping the second somatosensoryarea in humans. Somatosens. Motor Res. 10:297–308.

Coghill, R. C., Talbot, J. D., Evans, A. C., Meyer, E., Gjedde, A.,Bushnell, M. C., and Duncan, G. H. (1994). Distributed processingof pain and vibration by the human brain. J. Neurosci. 14:4095–4108.

Costanzo, R. M., and Gardner, E. P. (1980). A quantitative analysis ofresponses of direction-sensitive neurons in somatosensory cortex ofawake monkeys. J. Neurophysiol. 43:1319–1341.

Craig, A. D., Krout, K., and Zhang, E.-T. (1995). Cortical projectionsof Vmpo, a specific pain and temperature relay in primate thala-mus. Soc. Neurosci. Abstr. 21:1165.

Davis, K. D., Wood, M. L., Crawley, A. P., and Mikulis, D. J. (1995).fMRI of human somatosensory cortex during painful electricalnerve stimulation. Neuroreport 7:321–325.

Drevets, W. C., Burton, H., Videen, T. O., Snyder, A. Z., Simpson, J. R.,and Raichle, M. E. (1995). Blood flow changes in human somatosen-sory cortex during anticipated stimulation. Nature 373:249–252.

Duffy, F. H., and Burchfiel, J. L. (1971). Somatosensory system:organizational hierarchy from single units in monkey area 5.Science 172:273–275.

Flament, D., Ellermann, J. M., Kim, S., Ugurbil, K., and Ebner, T. J.(1996). Functional magnetic resonance imaging of cerebellar activa-tion during the learning of a visuomotor dissociation task. Hum.Brain Map. 4:210–226.

Forss, N., Hari, R., Salmelin, R., and Ahonen, A. (1994). Activation ofthe human posterior parietal cortex by median nerve stimulation.Exp. Brain Res. 99:309–315.

Fox, P. T., Burton, H., and Raichle, M. E. (1987). Mapping humansomatosensory cortex with positron emission tomography. J. Neuro-surg. 67:34–43.

Friedman, D. P., Murray, E. A., O’Neill, J. B., and Mishkin, M. (1986).Cortical connections of the somatosensory fields of the lateralsulcus of macaques: evidence for a corticolimbic pathway for touch.J. Comp. Neurol. 252:323–347.

Gallen, C. C., Sobel, D. F., Waltz, T., Aung, M., Copeland, B.,Schwartz, B. J., Hirschkoff, E. C., and Bloom, F. E. (1993).Noninvasive presurgical neuromagnetic mapping of somatosen-sory cortex. Neurosurgery 33:260–268.

Gallen, C. C., Schwartz, B. J., Rieke, K., Pantev, C., Sobel, D. F.,Hirschkoff, E. C., and Bloom, F. E. (1994). Intrasubject reliabilityand validity of somatosensory source localization using a largearray biomagnetometer. Electroenceph. Clin. Neurophysiol. 90:145–156.

Gallen, C. C., Schwartz, B. J., Bucholz, R. D., Ghaus, M., Barkley,G. L., Smith, J., Tung, H., Copeland, B., Bruno, L., Assam, S.,Hirschkoff, E. C., and Bloom, F. E. (1995). Presurgical localizationof functional cortex using magnetic source imaging. J. Neurosurg.82:988–994.

Hammeke, T. A., Zerrin, Y., Mueller, W. M., Morris, G. L., Haughton,V. M., Rao, S. M., and Binder, J. R. (1994). Functional magneticresonance imaging of somatosensory stimulation. Neurosurgery35:677–681.

Huttunen, J. (1986). Magnetic cortical responses evoked by tactilestimulation of the middle finger in man. Eur. J. Physiol. 407:129–133.

Huttunen, J., Hari, R., and Leinonen, L. (1987). Cerebral magneticresponses to stimulation of ulnar and median nerves. Electroen-ceph. Clin. Neurophysiol. 66:391–400.

Hyvarinen, J., and Poranen, A. (1978). Movement-sensitive anddirection and orientation-selective cutaneous receptive fields in thehand area of the post-central gyrus in monkeys. J. Physiol.(London) 283:523–537.

Iwamura, Y., and Tanaka, M. (1978). Postcentral neurons in handregion of area 2: Their possible role in the form discrimination oftactile objects. Brain Res. 150:662–666.

Iwamura, Y., Tanaka, M., Sakamoto, M., and Hikosaka, O. (1983).Converging patterns of finger representation and complex re-sponse properties of neurons in area 1 of the first somatosensorycortex of the conscious monkey. Exp. Brain Res. 51:327–337.

Jackson, G. D., and Duncan, J. S. (1996). MRI Neuroanatomy: A NewAngle on the Brain. Churchill Livingstone, Singapore.

Jenkins, W. M., Merzenich, M. M., Ochs, M. T., Allard, T., andGuic-Robles (1990). Functional reorganization of primary somato-sensory cortex in adult owl monkeys after behaviorally controlledtactile stimulation. J. Neurophys. 63:82–104.

Johansson, R. S., and Vallbo, Å. B. (1979). Tactile sensibility in thehuman hand: Relative and absolute densities of four types ofmechanoreceptive units in glabrous skin. J. Physiol. 286:283–300.

Kaas, J. H., Nelson, R. J., Sur, M., Lin, C.-S., and Merzenich, M. M.(1979). Multiple representations of the body within the primarysomatosensory cortex of primates. Science 204:521–523.

Karni, A., Gundela, M., Jezzard, P., Adams, M. M., Turner, R., andUngerleider, L. G. (1995). Functional MRI evidence for adult motorcortex plasticity during motor skill learning. Nature 377:155–158.

Kim, S. G., Ugurbil, K., and Strick, P. L. (1994). Activation of acerebellar output nucleus during cognitive processing. Science265:949–951.

Krubitzer, L. A., Clarey, J., Tweedale, R., Elston, G., and Calford,M. B. (1995). A redefinition of somatosensory areas in the lateralsulcus of macaque monkeys. J. Neurosci. 15:3821–3839.

Krubitzer, L. A., and Calford, M. B. (1992). Five topographicallyorganized fields in the somatosensory cortex of the flying fox:Microelectrode maps, myeloarchitecture, and cortical modules. J.Comp. Neurol. 317:1–30.

Krubitzer, L. A., and Kaas, J. H. (1990). The organization andconnections of somatosensory cortex in marmosets. J. Neurosci.10:952–974.

Lee, C. J., and Whitsel, B. L. (1992). Mechanisms underlyingsomatosensory cortical dynamics. I. In vivo studies. CerebralCortex 2:81–106.


Lin, W., Kuppusamy, K., Haacke, E. M., and Burton, H. (1996).Functional MRI in human somatosensory cortex activated bytouching textured surfaces. J. Magn. Res. Imag. 6:565–572.

Merzenich, M. M., Kaas, J. H., Sur, M., and Lin, C.-S. (1978). Doublerepresentation of the body surface within cytoarchitectonic areas3b and 1 in ‘‘SI’’ in the owl monkey (Aotus trivirgatus). J. Comp.Neurol. 181:41–74.

Merzenich, M. M., Nelson, R. J., Kaas, J. H., Stryker, M. P., Jenkins,W. M., Zook, J. M., Cynader, M. S., and Schoppmann, A. (1987).Variability in hand surface representations in areas 3b and 1 inadult owl and squirrel monkeys. J. Comp. Neurol. 258:281–296.

Mesulam, M.-M., and Mufson, E. J. (1982). Insula of the Old Worldmonkey. III. Efferent cortical output and comments on function. J.Comp. Neurol. 212:38–52.

Meyer, E., Ferguson, S. S. G., Zatorre, R. J., Alivisatos, B., Marrett,S., Evans, A. C., and Hakim, A. M. (1991). Attention modulatessomatosensory cerebral blood flow response to vibrotactile stimula-tion as measured by positron emission tomography. Ann. Neurol.29:440–443.

Mountcastle, V. B. (1984). Central nervous mechanisms in mechano-receptive sensibility. In Handbook of Physiology, the NervousSystem: Sensory Processes (J. M. Brookhart, V. B. Mountcastle, I.Darian-Smith, and S. R. Geiger, Eds.), Vol. 3, pp 789–878. WaverlyPress, Baltimore.

Mufson, E. J., and Mesulam, M.-M. (1982). Insula of the Old Worldmonkey. II. Afferent cortical input and comments on the claustrum.J. Comp. Neurol. 212:23–37.

Mufson, E. J., and Mesulam, M.-M. (1984). Thalamic connections ofthe insula in the rhesus monkey and comments on the paralimbicconnectivity of the medial pulvinar nucleus. J. Comp. Neurol.227:109–120.

Nicolelis, M. A., Baccala, L. A., Lin, R. C., and Chapin, J. K. (1995).Sensorimotor encoding by synchronous neural ensemble activity atmultiple levels of the somatosensory system. Science 268:1353–1358.

Nicolelis, M. A., Lin, R. C., Woodward, D. J., and Chapin, J. K. (1993).Induction of immediate spatiotemporal changes in thalamic net-works by peripheral block of ascending cutaneous information.Nature 361:533–536.

Ogawa, S., Lee, T. M., Kay, A. R., and Tank, D. W. (1990). Brainmagnetic resonance imaging with contrast dependent on bloodoxygenation. Proc. Natl. Acad. Sci. USA 87:9868–9872.

Ogawa, S., Tank, D. W., Menon, R., Ellermann, J. M., Kim, S.-G.,Merkle, H., and Ugurbil, K. (1992). Intrinsic signal changesaccompanying sensory stimulation: Functional brain mappingwith magnetic resonance imaging. Proc. Natl. Acad. Sci. USA89:5951–5955.

Okada, Y. C., Tanebaum, R., Williamson, S. J., and Kaufman, L.(1984). Somatotopic organization of the human somatosensorycortex revealed by neuromagnetic measurements. Exp. Brain Res.56:197–205.

Pardo, J. V., Fox, P. T., and Raichle, M. E. (1991). Localization of ahuman system for sustained attention by positron emission tomog-raphy. Nature 349:61–64.

Paul, R. L., Merzenich, M., and Goodman, H. (1972). Representationof slowly and rapidly adapting cutaneous mechanoreceptors of thehand in Brodmann’s areas 3 and 1 of Macaca mulatta. Brain Res.36:229–249.

Penfield, W., and Boldrey, E. (1937). Somatic motor and sensoryrepresentation in the cerebral cortex of man as studied by electricalstimulation. Brain 60:389–443.

Penfield, W., and Rasmussen, T. (1950). The Cerebral Cortex of Man:A Clinical Study of Localization of Function. MacMillan, New York.

Phillips, J. R., Johnson, K. O., and Hsiao, S. S. (1988). Spatial patternrepresentation and transformation in monkey somatosensory cor-tex. Proc. Natl. Acad. Sci. USA 85:1317–1321.

Powell, T. P. S., and Mountcastle, V. B. (1959). The cytoarchitecture ofthe postcentral gyrus of the monkey Macaca mulatta. Bull. JohnsHopkins Hosp. 105:108–131.

Prud’homme, M. J. L., Cohen, D. A. D., and Kalaska, J. F. (1994).Tactile activity in primate primary somatosensory cortex duringactive arm movements: Cytoarchitectonic distribution. J. Neuro-physiol. 71:173–181.

Puce, A. (1995). Comparative assessment of sensorimotor functionusing functional magnetic resonance imaging and electrophysiologi-cal methods. J. Clin. Neurophysiol. 12:450–459.

Raichle, M. E. (1994). Visualizing the mind. Sci. Am. 270:58–64.Ramsey, N. F., Tallent, K., Van Gelderen, P., Frank, J. A., Moonen,

C. T. W., and Weinberger, D. R. (1996). Reproducibility of human3D fMRI brain maps acquired during a motor task. Hum. BrainMap. 4:113–121.

Rao, S. M., Binder, J. R., Hammeke, T. A., Bandettini, P. A., Bobholz,J. A., Frost, J. A., Myklebust, B. M., Jacobson, R. D., and Hyde, J. S.(1995). Somatotopic mapping of the human primary motor cortexwith functional magnetic resonance imaging. Neurology 45:919–924.

Roland, P. E. (1981). Somatotopical tuning of postcentral gyrusduring focal attention in man. A regional cerebral blood flow study.J. Neurophysiol. 46:744–754.

Roland, P. E., Eriksson, L., Widen, L., and Stone-Elander, S. (1989).Changes in regional cerebral oxidative metabolism induced bytactile learning and recognition in man. Eur. J. Neurosci. 1:3–18.

Roland, P. E., Levin, B., Kawashima, R., and Åkerman, S. (1993).Three-dimensional analysis of clustered voxels in 15O-butanolbrain activation images. Hum. Brain Map. 1:3–19.

Seitz, R. J., Roland, P. E., Bohm, C., Greitz, T., and Stone-Elander, S.(1991). Somatosensory discrimination of shape: tactile explorationand cerebral activation. Eur. J. Neurosci. 3:481–492.

Seitz, R. J., and Roland, P. E. (1992). Vibratory stimulation increasesand decreases the regional cerebral bloodflow and oxidative metabo-lism: a positron emission tomography (PET) study. Acta Neurol.Scand. 86:60–67.

Sorenson, J. A., and Wang, X. (1996). Problems in estimatinghemodynamic response parameters from fMRI data. Hum. BrainMap. 4:265–272.

Stehling, M. K., Turner, R., and Mansfield, P. (1991). Echo-planarimaging: Magnetic resonance imaging in a fraction of a second.Science 254:43–50.

Suk, J., Ribary, U., Cappell, J., Yamamoto, T., and Llinas, R. (1991).Anatomical localization revealed by MEG recordings of the humansomatosensory system. Electroenceph. Clin. Neurophysiol. 78:185–196.

Sur, M. (1980). Receptive fields of neurons in areas 3b and 1 ofsomatosensory cortex in monkeys. Brain Res. 198:465–471.

Sur, M., Nelson, R. J., and Kaas, J. H. (1982). Representations of thebody surface in cortical areas 3b and 1 of squirrel monkeys:Comparisons with other primates. J. Comp. Neurol. 211:177–192.

Talairach, J. A., and Tournoux, P. (1988). Co-planar Stereotaxic Atlasof the Human Brain. Thieme Verlag, New York.

Turner, R., Jezzard, P., Wen, H., Dwong, K. K., LeBihan, D., Zeffiro,T., and Balaban, R. S. (1993). Functional mapping of the humanvisual cortex at 4 and 1.5 tesla using deoxygenation contrast EPI.Magn. Res. Med. 29:277–279.

Vogt, M. (1928). Uber omnilaminare Strukturdifferenzen und lineareGrenzen der architektonischen Felder der hinteren Zentralwind-ung des Menschen. J. Psychol. Neurol. 35:177–193.

282 GELNAR ET AL.

Whitsel, B. L., Dreyer, D. A., and Hollins, M. (1978). Representationof moving stimuli by somatosensory neurons. Fed. Proc. 37:2223–2227.

Whitsel, B. L., Dreyer, D. A., Hollins, M., and Young, M. G. (1980).The coding of direction of tactile stimulus movement: correlativepsychophysical and neurophysiological data. In Sensory Functionsof the Skin in Humans (D. Kenshalo, Ed.), pp 79–107. Plenum, NewYork.

Woolsey, C. N., Erickson, T. C., and Gilson, W. E. (1979). Localizationin somatic sensory and motor areas of human cerebral cortex asdetermined by direct recording of evoked potentials and electricalstimulation. J. Neurosurg. 51:476–506.

Yetkin, F. Z., Mueller, W. M., Hammeke, T. A., Morris, G. L., 3rd, andHaughton, V. M. (1995). Functional magnetic resonance imagingmapping of the sensorimotor cortex with tactile stimulation. Neuro-surgery 36:921–925.


fingertip representation in the human somatosensory cortex ... · primary somatosensory cortex...

Documents