FUNCTIONAL MAPPING OF THE HUMAN GLOBUS PALLIDUS: CONTRASTINGEFFECT OF STIMULATION IN THE INTERNAL AND EXTERNAL PALLIDUM IN

PARKINSON’S DISEASE

J. YELNIK,*† P. DAMIER,*‡ B. P. BEJJANI,‡ C. FRANCOIS,* D. GERVAIS,*‡ D. DORMONT,§ I. ARNULF,‡A. M. BONNET,‡k P. CORNU,¶ B. PIDOUX** and Y. AGID*‡ k

*INSERM U289, ‡Centre d’Investigation Clinique,kFederation de Neurologie and §Services de Neuroradiologie,¶Neurochirurgie and**Neurophysiologie, Hoˆpital de la Salpeˆtriere, 47 boulevard de l’Hoˆpital, F-75013 Paris, France

Abstract—Our objective was to elaborate a functional map of the globus pallidus by correlating the intrapallidal localization ofquadripolar electrodes implanted in parkinsonian patients with the clinical effect of the stimulation of each contact. Five patientswith l-DOPA-responsive Parkinson’s disease presenting severe motor fluctuations andl-DOPA-induced dyskinesias were treatedby continuous bilateral high-frequency stimulation of the globus pallidus. The effects of stimulation on parkinsonian disability weretested through each of the four stimulating contacts of each electrode. The anatomical localization of each of the stimulatingcontacts was determined by confronting the pre- and post-operative magnetic resonance imaging with the anatomical atlas ofSchaltenbrand and Wharen.34 The registration procedure comprised digitization of the atlas, the use of deformation tools to fit atlassections with magnetic resonance imaging sections, and three-dimensional reconstruction of both the atlas and the magneticresonance imaging sections. Analysis of the 32 stimulating contacts tested did not reveal a somatotopic organization in the pallidalregion investigated but demonstrated that high-frequency stimulation had contrasting effects depending on whether it was appliedto the external or the internal pallidum. Akinesia was improved by stimulation of the external pallidum but worsened by stimulationof the internal pallidum. In contrast, parkinsonian rigidity was improved by stimulation of either part of the pallidum. The areas inthe internal pallidum where stimulation worsened akinesia were those in which stimulation reduced or suppressedl-DOPA-induced dyskinesias. Conversely, stimulation applied to the external pallidum induced dyskinesias. The fact that rigidity wasimproved by stimulation of the internal and external pallidum suggests that the neuronal bases of parkinsonian rigidity are differentfrom those of akinesia and dyskinesias. The effect on akinesia and dyskinesias is in agreement with the current model of basalganglia circuitry10 if high-frequency stimulation activates rather than inhibits pallidal neurons, a possibility which is very likelysince there are marked anatomical, biochemical and electrophysiological differences between the globus pallidus and the sub-thalamic nucleus.

This study demonstrates that high-frequency stimulation of the globus pallidus in parkinsonian patients has contrasting effectsdepending on whether it is applied to the external or the internal part of this nucleus. The effect on akinesia and dyskinesias suggeststhat stimulation activates pallidal neurons, a result which challenges the generally accepted concept that high-frequency stimulationinactivates neurons in the region stimulated.q 2000 IBRO. Published by Elsevier Science Ltd. All rights reserved.

Key words:deep brain stimulation, basal ganglia, magnetic resonance imaging.

The last few years have seen a renewed interest in functionalsurgery as a treatment for Parkinson’s disease (PD), with aresurgence of pallidotomy3,22,27 and the use of deep brainstimulation (DBS) as an alternative, non-lesioning surgicaltherapy.5,6,35 In addition to progress in stereotactic surgery,this renewed interest is due to our improved knowledge ofthe functioning of the basal ganglia circuitry in physiologicaland pathological states. PD is characterized by a dopaminedeficiency in the striatum leading to an overactivity of theinternal globus pallidus (GPi) and the subthalamic nucleus(STN) in comparison with controls both in monkeys1,10,14,24,38

and humans.38 In accordance with the model, STN stimula-tion, which has been considered to act by neural inactivation,5

greatly improves all PD signs and allows patients’ drug treat-ment to be considerably reduced.25 In contrast, pallidotomy

and pallidal stimulation seem to have a less consistent effecton PD motor signs, ranging from moderate improve-ment3,21,12,21,37to absence of efficacy,17 but consistently reducel-DOPA-induced dyskinesias. This effect is difficult to recon-cile with the proposed model of basal ganglia circuitry, sincedyskinesias, which are considered to be the “mirror image” ofParkinsonism,10 are thought to result from a hypoactivity ofthe GPi. Lesions or inactivation of the GPi should thereforeworsen rather than improvel-DOPA-induced dyskinesias.

In a previous study, we investigated the paradox of pallidalsurgery by analysing the effects of high-frequency stimulationapplied to different parts of the globus pallidus (GP) in fivepatients with bilateral quadripolar pallidal electrodes.4 Stimu-lation in the dorsal part of the GP improved parkinsoniansigns and could generate dyskinesias, even in patients withoutl-DOPA. Stimulation in the posteroventral part of the GPsuppressedl-DOPA-induced dyskinesias but worsenedakinesia, thereby cancelling out the beneficial effects of thetreatment. We concluded that two different targets werepresent in the GP, implying that the dysfunction of two neuralnetworks is involved. Similar results have also beenreported.21 However, since magnetic resonance imaging(MRI) does not provide a clear identification of the contoursof the external globus pallidus (GPe) and GPi, the exact

Functional mapping of the human globus pallidus 77

77

NeuroscienceVol. 101, No. 1, pp. 77–87, 2000q 2000 IBRO. Published by Elsevier Science Ltd

Printed in Great Britain. All rights reserved0306-4522/00 $20.00+0.00PII: S0306-4522(00)00364-X

Pergamon

www.elsevier.com/locate/neuroscience

†Corresponding author. Tel.:133-142-16-06-32; fax:133-145-82-88-93.E-mail address:yelnik@ccr.jussieu.fr (J. Yelnik).Abbreviations:AC, anterior commissure; DBS, deep brain stimulation; GP,

globus pallidus; GPe, external part of the GP; Gpi, internal part of theGP; GPe/i, internal medullary lamina; MPTP, 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine; MRI, magnetic resonance imaging; PC,posterior commissure; PD, Parkinson’s disease; Pu/GP, external medul-lary lamina; STN, subthalamic nucleus; UPDRS, Unified Parkinson’sDisease Rating Scale.

localization of the stimulating contacts could not be deter-mined with sufficient accuracy. We thus decided to reanalyseour five patients with the aid of a more accurate method ofcontact localization. In this paper we do not report in detailthe clinical and surgical aspect of pallidal stimulation in PD,which has been done previously4,21 but we took advantage ofdifferent electrode orientation and placement in differentpatients to perform an accurate anatomofunctional correlationstudy. The goal was to analyse the anatomofunctional basis ofthe contrasting effects of GP stimulation by comparing theexact anatomical position of each contact of the implantedelectrodes with the clinical effects of stimulation deliveredthrough this contact.

EXPERIMENTAL PROCEDURES

Patients

Five patients, already tested in our previous study,4 aged from 44 to65 years (mean 55), with bilateral quadripolar electrodes implanted inthe GPi for the treatment of severe PD (mean disease duration 12^ 1years; mean Hoehn and Yahr score: 3.5^0.4 in the “off” drug condi-tion and 2.1 0.4 in the “on” drug condition) were included in thestudy. The target for the tip of the electrode was the posteroventral GPi.The electrodes were implanted according to our standard surgicalprocedure under MRI guidance, which allows trajectories with doubleobliquity to avoid the ventricles and cerebral blood vessels.4,12 Thestudy was approved by the local ethical committee and writteninformed consent was obtained from the PD patients. It was conductedin accordance with the Declaration of Helsinki.

Radiological tools

The day before surgery, stereotactic MRI was performed on a 1.5Tesla unit (Sigma General Electric) using an MRI-compatible LeksellG stereotactic frame (Elekta Instruments, Stockholm, Sweden), todetermine the site and trajectory for the implantation of the quadri-polar electrodes. Stereotactic MRI acquisition consisted of a three-dimensional spoiled gradient recalled acquisition allowing us to obtaincontiguous 1.5-mm axial slices close to the horizontal plane of thebrain. The day after surgery, another MRI was performed in non-stereotactic conditions to visualize the final position of the electrodes.The absence of significant image distortion was previously demon-strated.12 Voxtool software (General Electric Europe, Buc, France),running on an Advantage Windows Workstation, allowed originalacquisitions to be reformatted in any desired plane of section (Fig. 1).12

MRI images were visualized on the computer monitor as three simul-taneous windows corresponding to three orthogonal planes.

The system provides a three-dimensional cursor that could be posi-tioned at any point of interest, which thus appeared simultaneously inthe three planes. Its coordinates were displayed with reference to aselected point. The mid-sagittal plane was first determined as beingthe plane containing the anterior commissure (AC) and posteriorcommissure (PC) points and a third point selected in the longitudinalcerebral fissure (falx cerebri). A line was traced between the AC andPC points (Fig. 1B). Using the Voxtool program, the projection of thisline was superimposed on each MR section. The PC point was selectedas the zero reference point. Thex-axis corresponded to the medio-lateral direction, they-axis to the anteroposterior direction and thez-axis to the dorsoventral direction.

Anatomical analysis

An anatomoradiological procedure was developed to determine theintracerebral localization of each contact with the highest possibleaccuracy. It proceeded in successive steps (Fig. 1).

Tracing of cerebral contours on the pre-operative magnetic reson-ance imaging.The pre-operative MRI was reformatted into horizontalsections parallel to the AC–PC line and perpendicular to the mid-sagittal plane. Sections containing the striatum (caudate nucleus andputamen) and the GP (GPe and GPi) were obtained every 1 mm, start-ing from 15 mm below AC–PC to 25 mm above AC–PC (Fig. 1A).For each patient, cerebral contours which gave rise to a distinct MRI

contrast, e.g. the cerebral cortex, caudate nucleus, putamen orventricles (Fig. 1C), were drawn on tracing paper on each MRI sectionand digitized using an XY digitizer (SummaSketchwIII, Summa-graphics, Seymoun, CT, USA). Each contour was traced separatelyand stored in a file corresponding to the section (Fig. 1D). A tracingprogram developed in our centre was used to superimpose the contoursof a given region obtained in successive sections. A three-dimensionalreconstruction of each region (striatum, GPe, GPi) was then obtained.This was used to verify that successive contours were smoothly alignedand to correct any imprecision of tracing. The internal and externalpallidal segments could not be identified directly from the MRIsections because their limits were not sharply visible on every section.

Tracing of the internal and external parts of the globus pallidus.Thecontours of the GPi and GPe were determined using the atlas ofSchaltenbrand and Wharen.34 Cerebral contours of the different nucleiforming the basal ganglia in the atlas were first digitized using the sameprocedure as that used for tracing cerebral contours in the pre-operativeMRI. The correspondence between the atlas sections and the MRIseries was then established for each patient. Two reference sectionswere used for this purpose, the section containing the AC–PC line andthe section containing the upper border of the putamen, these land-marks being clearly identifiable in both MRI and atlas sections. Alinear transformation of the atlas series was then performed alongthez-axis to fit these two reference level sections with the correspond-ing levels of the patient’s MRI sections. Each section of the atlas wasthen associated with a section of the 1-mm-spaced sections of thepatient’s MRI.

The sections of the atlas were then deformed so as to obtain the bestpossible registration between the atlas contours and the MRI contoursof each patient. This was done using a specific function of the tracingprogram with which all the contours of the atlas could be reduced orincreased in size. The deformation factors were different for thex- andy-axes but remained the same along the whole series of sections. Then,contours that were not visible on MRI sections, such as the external andinternal subdivisions of the GP, were transferred from the atlas andadded to the digitized section file of the patient (Fig. 1D). By perform-ing this procedure for every horizontal section of the MRI series, athree-dimensional cartography of the pallidal region of each patient,including the putamen, the external medullary lamina (Pu/GP), externalpallidum, internal medullary lamina (GPe/i) and internal pallidum, wasobtained.

Localization of the contacts on the post-operative magnetic reson-ance imaging.A post-operative MRI acquisition was performed tovisualize the contacts of the electrodes. Post-operative MR imageswere processed with the Voxtool program as for pre-operative MRI.Quadripolar electrodes were 1.3-mm-thick electrodes bearing four 1.5-mm-long contacts separated by 1.5-mm zones (distance between thecentre of consecutive contacts�3 mm). Each quadripolar electrodewas visualized in the oblique plane that contained its long axis. Anelectrode appeared as a hyposignal comprising a linear trace and fourcircular artefacts, 3–4 mm in diameter, corresponding to the fourstimulating contacts (Fig. 1E). The cursor provided by the Voxtoolsoftware was positioned at the centre of each contact. The three co-ordinates (x, y, z) of this position were given by the program withreference to the AC–PC system of coordinates, with the PC pointtaken as the zero point. To specify the localization of the contactswithin the GP, a horizontal (i.e. parallel to the AC–PC plane) MRsection passing through the cursor was generated (Fig. 1F) and thesame section of the pre-operative MR series was identified. The tracedcontours of this section (direct tracings plus contours transferred fromthe atlas) were then superimposed on to the horizontal post-operativesection and the intrapallidal position of the contact artefact determined(arrow in Fig. 1G). This horizontal section was considered as the mostreliable document, and it was from this that the localization of eachcontact was determined. Each contact was qualified as localized eitherin the GPe, in the GPi, at the putamino–pallidal junction (the Pu/GP) orat the junction between GPe and GPi (the GPe/i).

For illustrative purposes, the localization of the four contacts ofeach electrode was reported on one of the four frontal sections of theatlas illustrated in Fig. 2. Trajectories of electrodes were extrapolatedfrom this tracing. However, this representation is necessarily distortedsince it is an orthogonal projection on a single frontal section of elec-trodes that were obliquely oriented on both the frontal and sagittalplanes. Therefore, the contact localization illustrated in Fig. 2 does

J. Yelnik et al.78

not necessarily match the qualification determined on the horizontalsection appropriate to each contact.

Evaluation of the effect of stimulation through each electrode contacton motor disability

Ten days after implantation of the electrodes, the effects on motordisability of electrical monopolar stimulation through each of the fourcontacts of each electrode were assessed to determine the optimalcontact for subsequent therapeutic use and the parameters of stimula-tion. The evaluation procedure has been described elsewhere.4 In brief,evaluation was based on the different items of the Unified Parkinson’sDisease Rating Scale (UPDRS) part III.13 The UPDRS was performedwhile the patients had been withoutl-DOPA for at least 12 h. Theeffect of stimulation at each contact was analysed separately in a

randomized order. For the analysis of any given contact, motor evalu-ation was performed before stimulation was applied, in order to obtaina baseline UPDRS motor score. For each UPDRS item additional halfpoints were used to increase the sensitivity of the testing. Then, with aconstant frequency (130 Hz) and a pulse width of 60 s, the stimulationvoltage was incrementally increased in 0.5 V steps until a maximum of5 V, or less if disabling side-effects occurred. At every 0.5 V step,stimulation was applied for 5 min and a UPDRS motor score wasperformed. When stimulation had an effect on a UPDRS item, theeffect changed gradually as the voltage was increased, until a plateauwas reached (same score obtained for at least three different 0.5 Vsteps). If two or more plateaux were observed before 5 V, the lastone was considered for the analysis of the effect. The score for eachitem obtained during the plateau was termed the optimal score. Withreference to the baseline score, we calculated the percentage change

Functional mapping of the human globus pallidus 79

Fig. 1. Anatomic localization of the contacts of the stimulating electrodes. A series of sections parallel to the AC–PC horizontal plane is off-printed every1 mm of the pre-operative MRI acquisition (A). The line between the anterior (AC) and posterior (PC) commissures is defined on the midsagittal section (B)and is projected onto each section of the series. The main structures, cortex, putamen (Pu), globus pallidus (GP), are identified (C) and the whole section istraced on a digitizer (D). The GPi and GPe are identified by comparing the entire series of MRI sections with the entire series of horizontal sections in theSchaltenbrand and Wharen atlas (see Experimental Procedures). Each contact is identified on an oblique section containing the entire electrode and localizedby a three-dimensional cursor (arrow) (E). The horizontal section of the post-operative MRI acquisition, which contains the cursor (arrow) is off-printed in the

AC–PC system (F). The position of the contact is then traced on the digitized section (G).

J. Yelnik et al.80

AC- PC

3L4L3R

2R

1R1L

PC+23

PC+18 PC+15.5

PC+20

2L5L

5R

4R

D

M L

V

Fig. 2.

for each item. During the evaluation, the type (ballism, chorea ordystonia), location (upper or lower limbs, trunk, neck and face), andseverity (from 0 to 4: 0� absence of dyskinesia; 4� extremely severedyskinesia) of dyskinesias were also assessed. The evaluation study ofone contact took 2–3 h. Due to fatigue in some of the patients, we wereunable to perform the evaluation of all contacts in all patients. At leasttwo contacts per electrode were analysed, and a total of 32 out of 40contacts were analysed (Fig. 2). When the patient experienced dys-kinesias throughout the day after usual drug intake, we also evaluatedthe effects of stimulation on the abnormal involuntary movementswhen applied to each of these 32 contacts. Since the clinical statewas less stable in the “on” condition, we simply assessed whether ornot a 5 V stimulation applied for 5 min through a given contact wasable to significantly reduce or stopl-DOPA-induced dyskinesias.

RESULTS

Anatomical localization of electrode contacts

Identification of the anatomical contours of the pallidalregion. The correspondence between the sections of theatlas and the MRI sections of each patient was defined onthe basis of the distance between the most dorsal point ofthe putamen and the AC–PC plane, which was chosen as areliable index of the dorsoventral dimension. This index was16, 17, 11, 15 and 16 mm in the five patients (correspondingto the order given in Fig. 2, respectively) compared to 11 mmin the atlas. The AC–PC distance in the patients was 24.5,24.9, 26.0, 24.4 and 27.2 mm (same order) and 21.5 in theatlas. There was no significant correlation between the twoindices.

Localization of electrode contacts within the pallidalregion. The localization of the contacts was defined on thebasis of the three coordinates of each contact and by best-fitregistration of the postoperative MRI with the three-dimensionalanatomical contours obtained from the preoperative MR andthe atlas. This localization is represented in a series of fourfrontal sections of the atlas (Fig. 2).

In patient 1, the electrodes were aligned in a parasagittalplane. They were the most anterior of the series (Fig. 2,PC1 23). The left electrode was 7 mm more medial thanthe lateral one. It had one contact (the most dorsal) withinthe GPe, two in GPe/i and one in the GPi. The dorsal contactcould not be tested clinically. The right electrode had the mostventral contact within the GPe and the three dorsal contactswithin the Pu/GP. The uppermost contact appears to belocated in the putamen in Fig. 2 as it was located 2 mm rostralto the section illustrated (PC1 23). In patient 2, there was aslight difference (2 mm) in the anteroposterior positioning ofthe two electrodes, the right one being more anterior than theleft one. The right electrode had one contact in the GPe, twoin the GPe/i and one in the GPi. The left electrode had itsdorsal contact in the GPe and the three ventral contacts in theGPi. One contact in the right electrode and two in the left onewere not tested clinically (Fig. 2). In patient 3, the left and

right electrodes were close to each other (3 mm) and werecentrally located within the GP (Fig. 2, PC1 20). The twoventral contacts of each electrode were in the GPi while thetwo dorsal contacts were in the GPe. Two of the GPe contactswere not tested (Fig. 2). In patient 4, there was a big difference(4.5 mm) in the anteroposterior positioning of the two elec-trodes, the left one being more anterior than the right one (Fig.2). Each electrode had its dorsal contact in the GPe, onecontact in the GPe/i and the two ventral ones in the GPi(Fig. 2). All contacts were tested in this patient. In patient5, the two electrodes were located lateral to the GPi (Fig. 2,CP1 18). Two contacts were in the Pu/GP, two in the GPeand four in the GPe/i. Two of the latter contacts were nottested.

In summary, the localization of the 40 contacts was asfollows: five in the Pu/GP, 12 in the GPe, 10 in the GPe/iand 13 in the GPi. Of these 40 contacts, three in the GPe, threein the GPe/i and two in the GPi were not tested clinically. GPicontacts were located between the anteriority PC1 15.5 toPC1 23 (Fig. 2), i.e. in the anterior half of the GPi. GPecontacts were localized at the same frontal levels, i.e. in theanterior half but not at the anterior pole of the nucleus. In allcases the left electrode was located more medially than theright electrode of the same patient [difference (mean^S.E.M.)�4.2^1.7 mm].

Effects of stimulation through each electrode contact onmotor disability

The effects of stimulation applied through each of the 32contacts tested is presented in Fig. 3. The mediolateral axisbeing in abscissa and the dorsoventral axis in ordinates, eachcontact is represented by a square positioned according to itscoordinates. The effects of unilateral pallidal stimulation,whether GPe or GPi, were mainly observed in the contra-lateral part of the body.

Stimulation had different effects on upper limb akinesiawhen applied in different parts of the GP. Stimulation appliedthrough 13 of the 14 contacts in the GPe or Pu/GP improvedupper limb akinesia (mean�42%; range�9–87%); stimula-tion through one contact in the GPe, close to the GPe/i,worsened it. By contrast, stimulation through 11 of the 12contacts located in the GPi worsened contralateral upperlimb akinesia (mean�49%; range�6–320%), the onlycontact which improved it being located close to the GPe/i.In the GPe/i, two contacts improved, three worsened and onehad no effect on contralateral akinesia (Fig. 3). Despite smallvariations, results were similar when the different items ofakinesia (thumb, index tapping, opening and closing hand,repetitive and alternative pronosupination movement of thehand) were considered. Similar results were obtained forlower limb akinesia, although changes were less pronounced:improvement of contralateral lower limb akinesia when

Functional mapping of the human globus pallidus 81

Fig. 2. Anatomical positioning of the 40 contacts (five patients; two electrodes per patient; four contacts per electrode) determined by best-fit registration withthree-dimensional linear deformations of the Schaltenbrand and Wharen atlas (see Experimental Procedures). The upper left drawing is a three-dimensionalreconstruction of this atlas, showing the anteroposterior alignment of the successive contours of the caudate nucleus (Cd), putamen (Pu), externalpallidum(GPe) and internal pallidum (GPi). The horizontal axis is the horizontal AC–PC plane.X andYaxes are calibrated in millimeters. The four stimulating contactsof each electrode are shown on the most appropriate frontal level of the atlas, i.e. that which corresponds to the mean anteriority of the four contacts. The mostanterior level is at PC1 23 mm and the most posterior is at PC1 15.5 mm. Atlas sections are frontal sections (D�dorsal; V ventral; M�medial;L� lateral). The trajectory of each electrode is indicated by a line and identified by the patient number (1–5) and the side (R for right, L for left). Allcontacts are represented by circles, the black circles indicating the eight contacts not tested clinically. Note that this representation is an approximation of theactual location of the 40 contacts, which are in fact localized at different anteroposterior levels (for example, the three uppermost contacts of electrode 1R werelocated in the external medullary lamina). The distance between successive contacts is not constant because the obliquity of the different electrodes differed.

stimulation was applied through 11 of the 14 contacts locatedin the GPe and Pu/GP (mean 52%; range�9–87%) andworsening when stimulation was applied through six of the12 contacts located in the GPi (mean�74%; range�17–250%). We observed either no effect on ipsilateral limbs oronly a slight modification comparable to that observed on the

contralateral side. Effects on gait impairment were similar tothose obtained for lower limb akinesia (Fig. 3). The moderaterest tremor observed in three patients was improved whenstimulation was applied through four contacts localized inthe GPe and through three in the GPe/i zone. Only a slightimprovement was noted when stimulation was applied

J. Yelnik et al.82

Fig. 3. Effect of high-frequency stimulation through 32 intrapallidal contacts on the main motor features of PD (upper limb and lower limb akinesia, gaitdisturbance and rigidity). Each of the 32 contacts tested in this study is represented by a square positioned according to its mediolateral (from 14 to24 mm inabscissa) and dorsoventral (from 6 to26 mm in ordinates) position. The mediolateral axis is traced at the level of the horizontal AC–PC plane (AC–PC). Theanteroposterior position of the contacts is not taken into account. The intrapallidal localization of each contact is indicated in the corresponding box: externalmedullary lamina (Pu/GP), external pallidum (GPe), internal medullary lamina (GPe/i), and internal pallidum (GPi). The intensity of shading corresponds to

the clinical effect of high-frequency stimulation through each contact: red�worsening; green� improvement.

through four contacts located in the GPi (Table 1). For theother items of the UPDRS III (speech, facial expression, arisefrom chair, posture, postural stability and body bradykinesia),there was a tendency to observe an improvement when stimu-lation was applied in the GPe and a worsening when stimula-tion was applied in the GPi, but no clear picture ofsomatotopic organization of the GP could be drawn fromthese results (Table 1).

The improvement of contralateral upper and lower limbrigidity was homogenous (mean 59%; range�25–100%)whether stimulation was applied in the Pu/GP, GPe, GPe/ior GPi. Only three contacts (one in the GPe, one in theGPe/i, one in the GPi) had no effect on rigidity. The intensityof the effect of stimulation on rigidity was similar in allinvestigated parts of the GP (Fig. 3). There was no clearsomatotopic distribution of the effect (upper limb versuslower limb). The effect of stimulation was less consistent onneck rigidity (improvement with 15/32 contacts). Stimulationhad no effect on ipsilateral rigidity.

Pallidal stimulation induced abnormal involuntary move-ments when a high voltage (at least higher than the optimalvoltage to change akinesia) was, used through some contacts.The most frequent type ofl-DOPA-induced dyskinesia wasdystonia affecting the hemibody contralateral to the stimula-tion, whether or not restricted to a limb and sometimes affect-ing neck, back or face. Dystonia was observed afterstimulation was applied through contacts localized either inthe Pu/GP, the GPe, the GPe/i zone or the GPi with no cleartopographic preference (Fig. 4). Stimulation also inducedchoreo-ballic movements in the hemibody contralateral tothe stimulation (Fig. 4). These movements were usuallyintense and similar to the movements experienced by thepatient while under the effect of antiparkinsonian drugs.They were only observed with stimulation applied throughcontacts localized in the GPe (3 contacts), the Pu/GP or theGPe/i zone (two contacts).

When the patient experiencedl-DOPA-induced dyskine-sias, the latter were significantly reduced or abolished whenstimulation was applied through contacts localized in theventrolateral GP (i.e. eight contacts in the GPi, three in theGPe/i, and only one in the ventral part of the GPe) (Fig. 4).

In the five patients, the contact that was used clinically

for each stimulated side was chosen in order to reach acompromise between improvement of akinesia by GPe stimu-lation and reduction ofl-DOPA-induced dyskinesias by GPistimulation.

DISCUSSION

The goal of this study was to obtain a functional map of thehuman GP by relating the clinical effects of high-frequencymonopolar stimulation to precise anatomical localizationswithin the GP. As the oblique trajectory of the electrodeswas adjusted to each patient and as the electrodes werequadripolar, different locations could be tested in differentPD patients. The anatomical and clinical studies were carriedout independently. The anatomists determined the locationof each contact without knowing the clinical effect of stimu-lation applied at this level and the clinicians did not knowthe exact anatomical localization of the contacts they weretesting.

In the five patients studied the stimulating electrodes werelocalized in the anterior half of the GPe and GPi. This local-ization is more anterior than the posteroventral target definedfor pallidotomy,22 but it has been shown that the sensorimotorpart of the GPi comprises the caudal two-thirds of the nucleusin the monkey.16 A comparison with Fig. 5 of Ref. 16 indi-cates that only electrodes 1L and 3L and the dorsal contact of2L could be localized in the associative part of the GP, theseven remaining electrodes being in the sensorimotor part.One electrode in patient 1 and two electrodes in patient 5were more lateral than expected and therefore had no contactsin the GPi.

An analysis of the 32 stimulating contacts tested did notreveal a somatotopic organization in the investigated pallidalregion but demonstrated that high-frequency stimulation hadcontrasting effects depending on whether it was applied to theGPe or the GPi. Although this study did not cover the entireGP, it comprised the sites in which parkinsonian motor signsare known to be corrected, i.e. various parts of the sensori-motor GP, a region in which a somatotopic organization hasbeen described in the monkey.11,42 The monkey GP is,however, a region in which information from sensory andmotor cortices converges on small functional sites.15 From

Functional mapping of the human globus pallidus 83

Table 1. Effect of high-frequency stimulation on Parkinson’s disease signs (different items of the Unified Parkinson’s Disease Rating Scale part III) whenapplied in different parts of the globus pallidus

Pu/GP (5) GPe (9) GPe/GPi (6) GPi (12)

Resting tremor:contralateral limbs 111 1 (3) 111 (2) 111 1 (2)

2 (1) 22 (2)face, lips, chin 1 (1) 2 (1) 11 (1)

2 (1) 22 (1)Speech 111 1 (2) 1 (1) 222 (2)

1 (1)Facial expression 2 (1) 111 1 (2) 1 (1) 2222 (2)

1 (1)Arise from chair 11 (1) 2 (1) 2222 (3)Posture 11111 (3) 22 (2) 222 (2)

1 (1)Postural stability 11 (2) 2 (1) 22 (1)Body bradykinesia 11111 (3) 1 (1) 22 (1)

Pu/GP�external medullary lamina (five contacts); GPe external part of the GP (nine contacts); GPi� internal part of the GP (12); GPe/GPi� internalmedullary lamina (six contacts).1 denotes improvement and2 denotes worsening. The number of signs (1 or 2) indicates the number of contacts throughwhich stimulation (, 5 V) induced the effect. In parentheses, the number of patients in whom the effect was observed.

the convergent model suggested by this study, a clear somato-topic organization is not expected in the GP, at least inmonkeys. Moreover, our mapping study was conducted inPD patients, in whom the somatotopic organization is prob-ably altered, in line with results obtained in monkeys renderedparkinsonian by 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine(MPTP) intoxication.36

Technical considerations

Anatomical localization of the electrode contacts within thepallidal region. There is no simple means of determiningdirectly the exact anatomical position of the contacts in thebrain of each treated patient. Indeed, although MRI providesdetailed images of the living brain, not all cerebral structuresare clearly discernible. Furthermore, DBS electrodes inducelarge artefacts that make visualization of the four contactsdifficult. For this reason, we were obliged to develop a proce-dure that combined pre-operative and post-operative MRIsand registration with an anatomical atlas.

The anatomical boundaries of the pallidal region were iden-tified on pre-operative MRI, which was performed understereotactic conditions. The Voxtool program was used toreformat the MRI in the AC–PC system of reference and tolocalize each voxel of the image with a precision of 1 mmwith reference to this conventional stereotactic system.12 TheT1-weighted sequence acquisition used in this study providedcontrasted images on which the ventricles, the caudatenucleus and the putamen could easily be delineated. As thetwo parts of the GP were not clearly discernible on the wholeseries of MRI sections, they were delineated by registrationof the serial MRI sections with the histological atlas of

Schaltenbrand and Wharen.34 This registration procedureencountered three technical problems. Firstly, the Schaltenbrandatlas does not provide regularly interspaced sections and notevery section contains the entire caudate nucleus, putamenand brainstem contours, which are good registration landmarks.Secondly, the contours of the atlas were always more medialthan the corresponding contours in the living brain due toshrinkage of the third ventricle. This was best corrected byaligning the lateral limit of the third ventricle of the MRIimage with the midline of the atlas. Thirdly, the rules to beapplied to register one brain with another are not completelyknown. In this study, we used linear deformations of the atlas inorder to reach a best fit with MRI sections. We applied inde-pendent factors along the mediolateral, dorsoventral and antero-posterior axes because size variations along the three axes arenot correlated, as shown in the monkey.33 It is not known,however, whether brain variations in a given direction are thesame for all cerebral structures or if different rules apply to eachnucleus or bundle. Despite these uncertainties, taking intoaccount the three-dimensional structure of the GP allowed usto determine the limit between the GPe and the GPi.

The centre of each contact was localized on the post-operative MR. We considered the centre of the artefact asthe centre of the contact, though it has not yet been demon-strated that the artefact is homogenous around the contact.The two methods used to localize the contacts (measurementof coordinates with reference to the AC–PC system of co-ordinates and anatomical mapping based on the registration ofthe postoperative MRI with the pre-operative MRI and thedigitized atlas) were concordant: the most dorsolateralcontacts were localized in the putamen or the GPe and themost ventromedial ones were localized in the GPi (Fig. 3). As

J. Yelnik et al.84

Fig. 4. Effect on dyskinesias of high-frequency stimulation through 32 intrapallidal contacts. Same representation as in Fig. 3. The diagram on the left showsdyskinesias induced by stimulation of the different contacts: dystonia is in blue, choreoballism in yellow. The diagram on the right shows the contacts through

which stimulation decreases or abolishesl-DOPA-induced dyskinesias.

the contacts in our patients were localized along obliquetrajectories, it was not possible to represent them on a singlemap without distortion. In Fig. 2, the contacts are representedat four different rostrocaudal levels of the GP, whichpreserves the mean anteriority of each electrode. In Fig. 3,the anteriority of the contacts is not considered and eachcontact is represented by a square in which its anatomicallocalization is indicated and a colorimetric code indicatesthe clinical effects of a stimulation applied through it.

Effects of stimulation through each electrode contact onmotor disability.For a given PD sign, the effect of stimulationgradually changed when voltage increased until a plateau wasreached. We only considered the effect obtained with voltagesbelow 5 V. Although the exact volume through which stimu-lation exerts an action is not known, it is generally acceptedthat at less than 5 V the effect of stimulation is restricted to afew millimetres around the stimulated contact.1,5,26 The factthat stimulation can have an opposite effect on akinesia whenapplied through two contacts 1.5 mm apart on the same elec-trode, supports this postulate. In a few cases, two plateauxwere observed before a 5 V voltage was reached but the effectof stimulation at these two plateaux was consistent, i.e.improvement or worsening, with the effect at the secondplateau being more pronounced. To increase the contrastbetween clinical effects, the score obtained during the secondplateau was considered.

Anatomical basis for the contrasting effects of pallidalstimulation

The segregation of pallidofugal fibres into the ansa lenti-cularis and the lenticular fasciculus has been proposed toexplain the opposite effect obtained by DBS in the ventraland dorsal parts of the GP.21 These two bundles constitutethe main efferent pathway by which the GPi projects to themotor thalamus. They arise from different parts of the GPi andfollow different trajectories.31 Fibres of the ansa lenticularisarise mainly from the lateral portion of the GPi, course ante-riorly and ventrally, cross the anterior part of the posteriorlimb of the internal capsule and then curve posteriorly to enterForel’s field H. Fibres of the lenticular fasciculus arise fromthe medial portion of the GPi, course medially and dorsally,cross the posterior limb of the internal capsule slightly caudalto the ansa lenticularis and rostral to the STN and curveposteriorly to form Forel’s field H2 dorsal to the STN. Fibresof the lenticular fasciculus and ansa lenticularis both reachForel’s field H and curve rostrally and dorsally to form thethalamic fasciculus or Forel’s field H1 located dorsal to thezona incerta. Thus, although they follow different trajectoriesat their origin, fibres of the two systems finally join into asingle system, the thalamic fasciculus, which conveys pallidaloutput to the motor thalamus. Can these two efferent bundlesconvey different information? In parkinsonian patients, adifference in spontaneous firing rate has been reported betweenneurons of the inner portion of the GPi, which displayed anoveractivity that could contribute to bradykinesia, andneurons of the outer portion, which had the same activity asthat of neurons of the GPe.20 This observation would suggestthat the ansa lenticularis and lenticular fasciculus transmitcontrasting information to the thalamocortical system, butthis is hard to reconcile with the fact that both portions ofthe GPi receive the same striatal input and both project onto

the entire dorsoventral extent of the motor thalamus in themonkey.2 Moreover, studies in monkeys based on retrogradetransneuronal transport of herpes simplex virus19 and on elec-trophysiological identification of pallidal neurons by corticalstimulation42 showed that neurons of the GPi are not segre-gated according to the inner/outer subdivision but organizedinto different dorsoventral channels connected to the supple-mentary motor area, the premotor cortex and the motorcortex. Thus, it is not clearly demonstrated that the ansa lenti-cularis and the lenticular fasciculus are the anatomical basisfor the contrasting effects observed after DBS in the GP.

The results of the present study suggest that the contrastingclinical effect observed in the GP is in fact due to the local-ization of ventral contacts in the GPi and dorsal contacts in theGPe. This result is in keeping with the three-dimensionalanatomy of the GP. Indeed, the three-dimensional shape ofthe GPi is not pyramidal as suggested by its appearance onfrontal sections (Fig. 2), but is shaped rather like a curvedbanana flattened along its long axis. This shape is most clearlyvisible on horizontal sections (see Fig. 1D). The width of thestructure, which is almost constant along its entire extent, isaround 4 mm. As there was a 3 mm interval between thecentre of two consecutive contacts and as the electrode trajec-tory was almost perpendicular to the long axis of the GPi, thisimplies that a maximum of two contacts, the most ventralones, could be located within the GPi, the remaining dorsalones being within the GPe. Thus, the anatomical characteris-tics of the GP and the length covered by the four contactsstrengthens the conclusion that the dorsal contacts of thequadripolar electrodes were actually in the GPe and not inthe GPi. This anatomical explanation is in agreement with thefact that the GPe and GPi are parts of the indirect and directpathways which are known to have opposite effects on basalganglia output.1,10 However, if high-frequency stimulationacts through neuro-inactivation of the stimulated target, assuggested by thalamic25 and STN6 stimulation, these resultsare in contradiction with the current model of basal gangliacircuitry.10 A first explanation could be that the model is infact inadequate or incomplete as suggested recently.8,23,28,30,32

A second explanation could, be attributed to anatomicaldifferences between the GP and the STN. The STN is asmall nucleus (13× 6× 3 mm) containing a high density ofneurons with highly ramified and relatively restricted dendriticarborizations (1200× 600× 300mm)40 whereas the GPe andGPi are large nuclei (GPe�25× 13× 4 mm; GPi�18× 8×4 mm) comprising a low density of neurons that havefew branched dendrites and larger dendritic arborizations(1500× 1000× 250mm).41 Therefore, stimulation is likelyto act upon most STN neurons and thus on the activity ofthe entire STN, but on only a restricted part of the GP. Thiscould explain why STN stimulation has a beneficial effect onthe three parkinsonian symptoms25 whereas the results ofpallidal stimulation are much more variable.4,21,37 However,the exact mechanism by which high-frequency stimulationacts on these nuclei is still unclear. Stimulation could involve,to varying, degrees, afferent fibres, cell bodies, efferent axonsand local axon collaterals. In addition, its effect could bedifferent on excitatory glutamatergic neurons of the STNand inhibitory GABAergic neurons of the GP.

Effect of pallidal stimulation on parkinsonian symptoms

Improvement of l-DOPA-induced dyskinesias by GPi

Functional mapping of the human globus pallidus 85

stimulation could be explained by experimental studies.Matsumura and colleagues29 induced dyskinesias by localinjections of bicuculline in the GPe of normal monkeys and,contrary to the expected hyperactivity of the entire GPe andhypoactivity of the GPi, they recorded groups of activatedneurons surrounded by groups of inhibited neurons insideeach pallidal segment. They concluded that dyskinesias didnot result from an imbalance between the two pallidalsegments but rather from an imbalance between neuronsinside each pallidal segment. Confirmation that dyskinesiasare not induced by a hyperactivity of the GPe was provided byBlanchet and colleagues,7 who demonstrated that lesion of theGPe did not relievel-DOPA-induced dyskinesias in MPTP-treated monkeys. These data suggest that dyskinesias cannotbe explained by an overall hypo- or hyperactivity of a givenpallidal segment. A more subtle dysfunctioning of the internalcircuitry of each nucleus of the basal ganglia is more likely tobe responsible for inducing dyskinesias. A plausible explana-tion is that high-frequency stimulation of the GPi corrects thisdysfunction by imposing a more regular neuronal activity onbasal ganglia output.

Rigidity was improved wherever pallidal stimulation wasapplied in the external or internal parts of the nucleus. Themost surprising result is the concomitant improvement ofrigidity and worsening of akinesia when stimulation wasapplied in the sensorimotor part of the GPi, as alreadyobserved.4,21 A possible explanation could be that high-frequency stimulation of the GPi acts on different neuralnetworks, one improving rigidity and the other worseningakinesia. For example, akinesia could be related to dysfunc-tion of the pallido–thalamic projections while rigidity couldinvolve dysfunction of pallidal projections to the pedunculo-pontine nucleus. As 86% of GPi neurons project to both thethalamus and the pedunculopontine nucleus,18 this explana-tion does, however, seem rather unlikely. A definitive

interpretation of this paradox is difficult to formulate givenour insufficient knowledge of the exact mechanism of con-tinuous high-frequency stimulation at the neuronal level.

Worsening of akinesia by stimulation of the GPi supposes areinforcement of the inhibitory output of the GPi to the thala-mus, thus an activation of GPi inhibitory neurons. This holdstrue for improvement of akinesia by stimulation of the GPe.This suggests that, contrary to what is observed in the STN6

and thalamus,25 high-frequency stimulation could act throughneuro-activation in the GP. This possibility is in fact verylikely. Indeed, many neurons in the GPe and GPi dischargecontinuously at high frequencies9,11,14 and could thereforefollow high-frequency stimulation. Conversely, thalamicneurons39 and STN11 discharge at such high rates only duringshort bursts and could be saturated and inhibited by high-frequency stimulation.

CONCLUSIONS

This anatomo-clinical study demonstrates that high-frequency stimulation has contrasting effects depending onwhether it is applied to the GPe or the GPi. These resultsagree with the current model of the basal ganglia circuitry10

if high-frequency stimulation activates rather than inhibitsGPe and GPi neurons. This possibility is very likely sincethere are major anatomical (low versus high neuronaldensity), biochemical (GABAergic versus glutamatergiccontent) and electrophysiological (continuous versus burstdischarge) differences between the GP and the STN. Theseresults challenge the generally accepted concept that high-frequency stimulation inactivates neurons in the regionstimulated.

Acknowledgements—This study was supported by INSERM (AO CIC1997), PHRC 1996, and the National Parkinson Foundation.

REFERENCES

1. Albin R. L., Young A. B. and Penney J. B. (1989) The functional anatomy of basal ganglia disorders.Trends Neurosci.12, 366–375.2. Arecchi-Bouchhioua P., Yelnik J., Franc¸ois C., Percheron G. and Tande´ D. (1997) Three-dimensional morphology and distribution of pallidal axons

projecting to both the lateral region of the thalamus, and the central complex in primates.Brain Res.754,311–314.3. Baron M. S., Vitek J. L., Bakay R. A., Green J., Kaneoke Y., Hashimoto T., Turner R. S., Woodard J. L., Cole S. A., McDonald W. M. and DeLong M. R.

(1996) Treatment of advanced Parkinson’s disease by posterior GPi pallidotomy: l-year results of a pilot study.Ann. Neurol.40, 355–366.4. Bejjani B. P., Damier P., Arnulf I., Bonnet A. M., Vidailhet M., Dormont D., Pidoux B., Cornu P., Marsault C. and Agid Y. (1997) Pallidal stimulation for

Parkinson’s disease—Two targets?Neurology49, 1564–1569.5. Benabid A. L., Pollak P., Gervason C., Hoffmann D., Gao D. M., Hommel M., Perret J. E. and De Rougemont J. (1991) Long-term suppression of tremor

by chronic stimulation of the ventral intermediate thalamic nucleus.Lancet337,403–406.6. Benabid A. L., Pollak P., Louveau A., Henry S. and De Rougemont J. (1987) Combined (thalamotorny and stimulation) stereotactic surgery of the VIM

thalamic nucleus for bilateral Parkinson disease.Appl. Neurophysiol.50, 344–346.7. Blanchet P. J., Boucher R. and Be´dard P. (1994) Excitotoxic lateral pallidotomy does not relievel-DOPA-induced dyskinesia in MPTP parkinsonian

monkeys.Brain Res.650,32–39.8. Chesselet M. F. and Delfs J. M. (1996) Basal ganglia and movement disorders: an update.Trends Neurosci.19, 417–422.9. DeLong M. R. (1971) Activity of pallidal neurons during movement.J. Neurophysiol.34, 414–427.

10. DeLong M. R. (1990) Primate models of movement disorders of basal ganglia origin.Trends Neurosci.13, 281–285.11. DeLong M. R., Crutcher M. D. and Georgopoulos A. P. (1985) Primate globus pallidus and subthalamic nucleus: functional organization.J. Neuro-

physiol.53, 530–543.12. Dormont D., Cornu P., Pidoux B., Bonnet A. M., Biondi A., Oppenheim C., Hasboun D., Damier P., Cuchet E., Philippon J., Agid Y. and Marsault C.

(1997) Chronic thalamic stimulation with three-dimensional MR stereotactic guidance.Am. J. Neuradiol.18, 1093–1107.13. Fahn S. and Elton R. (1987) Unified Parkinson’s disease rating scale. InRecent Developments in Parkinson’s disease(eds Fahn S., Marsden C. D., Calne

D. B. and Goldstein M.). Macmillan Health Care Information, Florham Park, NJ.14. Filion M. and Tremblay L. (1991) Abnormal spontaneous activity of globus pallidus neurons in monkeys with MPTP-induced Parkinsonism.Brain Res.

547,142–151.15. Flaherty A. W. and Graybiel A. M. (1993) Output architecture of the primate putamen.J. Neurosci.13, 3222–3237.16. Franc¸ois C., Yelnik J., Percheron G. and Fenelon G. (1994) Topographic distribution of the axonal endings from the sensorimotor and associative

striatum in the macaque pallidum and substantia nigra.Expl Brain Res.102,305–318.17. Friedman J. H., Epstein M., Sanes J. N., Lieberman P., Cullen K., Lindquist C. and Daamen M. (1996) Gamma knife pallidotomy in advanced

Parkinson’s disease.Ann. Neurol.39, 535–538.18. Harnois C. and Filion M. (1982) Pallidofugal projections to thalamus and midbrain: a quantitative antidromic activation study in monkeys and cats.Expl

Brain Res.47, 277–285.

J. Yelnik et al.86

19. Hoover J. H. and Strick P. L. (1993) Multiple output channels in the basal ganglia.Science259,819–821.20. Hutchison W. D., Lozano A. M., Davis K. D., Saint-Cyr J. A., Lang A. E. and Dostrovsky J. O. (1994) Differential neuronal activity in segments of the

globus pallidus in Parkinson’s disease patients.NeuroReport5, 1533–1535.21. Krack P., Pollak P., Limousin P., Hoffman D., Benazzouz A., LeBas J. F., Koudsie A. and Benabid A. L. (1998) Opposite motor effects of pallidal

stimulation in Parkinson’s disease.Ann. Neurol.43, 180–192.22. Laitinen L. V., Bergenheim A. T. and Haziz M. I. (1992) Leksell’s posteroventral pallidotomy in the treatment of Parkinson’s disease.J. Neurosurg.76,

53–61.23. Levy R., Hazrati L. N., Herrero M. T., Vila M., Hassani O. K., Mouroux M., Ruberg M., Asensi H., Agid Y., Fe´ger J., Obeso J. A., Parent A. and Hirsch E.

C. (1997) Re-evaluation of the functional anatomy of the basal ganglia in normal and parkinsonian states.Neuroscience76, 335–343.24. Levy R., Herrero M. T., Ruberg M., Villares J., Faucheux B., Guridi J., Guillen J., Luquin M. R., Javoy-Agid F., Obeso J. A., Agid Y. and Hirsch E. C.

(1995) Effects of nigrostriatal denervation and L-dopa therapy on the GABAergic neurons of the striatum in MPTP-treated monkeys and Parkinson’sdisease: Anin situ hybridization study of GAD(67) mRNA.Eur. J. Neurosci.7, 1199–1209.

25. Limousin P., Krack P., Pollak P., Benazzouz A., Ardouin C., Hoffmann D. and Benabid A. L. (1998) Electrical stimulation of the subthalamic nucleusinadvanced Parkinson’s disease.New Engl. J. Med.339,1105–1111.

26. Limousin P., Pollak P., Benazzouz A., Hoffmann D., Le Bas J. F., Broussolle E., Perret J. E. and Benabid A. L. (1995) Effect of parkinsonian signs andsymptoms of bilateral subthalamic nucleus stimulation.Lancet345,91–95.

27. Lozano A. M., Lang A. E., Galvez-Jimenez N., Miyasaki J., Duff J., Hutchinson W. D. and Dostrovsky J. O. (1995) Effect of GPi pallidotomy on motorfunction in Parkinson’s disease.Lancet346,1383–1387.

28. Marsden C. D. and Obeso J. A. (1994) The functions of the basal ganglia and the paradox of stereotaxic surgery in Parkinson’s disease.Brain 117,877–897.

29. Matsumura M., Tremblay L., Richard H. and Filion M. (1995) Activity of pallidal neurons in the monkey during dyskinesia induced by injection ofbicuculline in the external pallidum.Neuroscience65, 59–70.

30. Obeso J. A., Rodriguez M. C. and DeLong M. R. (1997) Basal ganglia pathophysiology. A critical review. InThe Basal Ganglia and New SurgicalApproaches for Parkinson’s Disease(eds Obeso J. A., DeLong M. R., Ohye C. and Marsden C. D.). Lippincott-Raven, Philadelphia.

31. Parent A. (1996) Carpenter’s Human Neuroanatomy, Williams & Wilkins, Baltimore.32. Parent A. and Ciechetti F. (1998) The current model of basal ganglia organization under scrutiny.Mov. Disord.13, 199–202.33. Percheron G. (1975) Ventricular landmarks for thalamic stereotaxy in Macaca.J. Med. Primatol.4, 217–244.34. Schaltenbrand G. and Wharen W. (1977) Atlas for Stereotaxy of the Human Brain, Georg thieme Verlag, Stuttgart.35. Siegfried J. and Lippitz B. (1994) Bilateral chronic electrostimulation of ventroposterolateral pallidum: a new therapeutic approach for alleviating all

parkinsonian symptoms.Neurosurgery35, 1126–1130.36. Tremblay L., Filion M. and Be´dard P. (1989) Responses of pallidal neurons to striatal stimulations in monkeys with MPTP-induced parkinsonism.Brain

Res.498,17–33.37. Tronnier V. M., Fogel W., Kronenbuerger M. and Steinvorth S. (1997) Pallidal stimulation: an alternative to pallidotomy?J. Neurosurg.87,700–705.38. Vila M., Levy R., Herrero M. T., Faucheux B., Obeso J. A., Agid Y. and Hirsch E. C. (1996) Metabolic activity of the basal ganglia in parkinsonian

syndromes in human and non-human primates: a cytochrome oxidase histochemistry study.Neuroscience71, 903–912.39. Vitek J. L., Ashe J., DeLong M. R. and Alexander G. E. (1994) Physiological properties and somatotopic, organization of the primate motor thalamus.

J. Neurophysiol.71, 1498–1513.40. Yelnik J. and Percheron G. (1979) Subthalamic neurons in primates: A quantitative and comparative analysis.Neuroscience4, 1717–1743.41. Yelnik L., Percheron G. and Franc¸ois C. (1984) A Golgi analysis of the primate globus pallidus. II. Quantitative morphology and spatial orientation of

dendritic arborizations.J. comp. Neurol.227,200–213.42. Yoshida S. I., Nambu A. and Jinnai K. (1993) The distribution of the globus, pallidus neurons with input from various cortical areas in the monkeys.Brain

Res.611,170–174.

(Accepted1 August2000)

Functional mapping of the human globus pallidus 87