chronic hyperkinetic condition of children with severe dystonia made it necessary to develop a short procedure that

J Neurosurg 96:673 679, 2002 Comparison of atlas- and magnetic resonance imaging based stereotactic targeting of the globus pallidus internus in the performance of deep brain stimulation for treatment of dystonia NATHALIE VAYSSIERE, PH.D., SIMONE HEMM, B.M.E., LAURA CIF, M.D., MARIE CHRISTINE PICOT, M.D., NINA DIAKONOVA, PH.D., HASSAN EL FERTIT, M.D., PHILIPPE FREREBEAU, M.D., AND PHILIPPE COUBES, M.D., PH.D. Departments of Pediatric Neurosurgery (Research Group on Movement Disorders in Children) and Biostatistics, University Hospital, Montpellier, France Object. To assess the validity of relying on atlases during stereotactic neurosurgery, the authors compared target coordinates in the globus pallidus internus (GPi) obtained using magnetic resonance (MR) imaging with those determined using an atlas. The targets were used in deep brain stimulation (DBS) for the treatment of generalized dystonia. Methods. Thirty-five patients, who were treated using bilateral DBS of the GPi, were included in this study. The target was selected on three-dimensional MR images by direct visual recognition of the GPi. The coordinates were automatically recorded using dedicated software. They were translated into the anterior commissure posterior commissure (AC PC) coordinate system by using a matrix transformation process. The same GPi target was defined, based on the locations of brain structures shown in the atlases of Schaltenbrand and Talairach. Magnetic resonance imaging based GPi target coordinates were statistically compared with the corresponding atlas-based coordinates by applying the Student t-test. A significant difference (p 0.001) was demonstrated in x, y, and z directions between MR imaging based and Schaltenbrand atlas derived target coordinates. The comparison with normalized Talairach atlas coordinates demonstrated a significant difference (p 0.01) in the y and z directions, although not in the x direction (p = 0.12). No significant correlation existed between MR imaging based target coordinates and patient age (p 0.1). No significant correlation was observed between MR imaging based target coordinates and patient sex in the y and z directions (p 0.9), although it was significant in the x direction (p 0.05). A significant variation in coordinates and the length of the AC PC line was revealed only in the y direction (p 0.005). Conclusions. A significant difference was found between target coordinates obtained by direct visual targeting on MR images (validated by postoperative clinical results) and those obtained by indirect targeting based on atlases. KEY WORDS globus pallidus stereotactic targeting magnetic resonance imaging E ARLY-onset generalized dystonia is a dyskinetic movement disorder that displays a poor response to pharmacological treatment. In the past, surgery has often been considered as a last resort for the most severe cases. Various targets have been selected for ablative or additive procedures (for example, GPi, thalamus, and subthalamic nucleus), but the results have often been incomplete and transient. More recently, DBS of the GPi has been proposed as an alternative approach, especially for use in children. 4,5,18,24 The young age, poor general state of health, and chronic hyperkinetic condition of children with severe dystonia made it necessary to develop a short procedure that Abbreviations used in this paper: AC = anterior commissure; BMFDRS = Burke-Marsden-Fahn Dystonia Rating Scale; DBS = deep brain stimulation; GPi = globus pallidus internus; MR = magnetic resonance; PC = posterior commissure; SD = standard deviation; 3D = three-dimensional. could be performed with the child in a state of general anesthesia and during which target localization was based only on 3D MR imaging, without concomitant intraoperative microelectrode recording. Such a procedure has been developed and validated at our center. 4,5,18,24 A strictly defined protocol allowed us to evaluate distortions of brain structures on MR images before and after surgery in daily practice, using the frame as a phantom. Many groups continue to use indirect targeting primarily, that is, coordinates calculated by measuring the distance between the target and an internal reference, for example the AC PC line. This line can be located with the aid of ventriculography, computerized tomography, or MR imaging. Distances are then deduced using a human brain atlas, hypothesizing that the proportions given in the atlas are applicable to any individual. It has long been known, however, that there are substantial individual variations in coordinates of subcortical nu- 673

N. Vayssiere, et al. FIG. 1. Upper Left: Axial transverse 3D spoiled-gradient MR image of the left basal ganglia obtained under stereotactic conditions. Upper Center: Map of the axial section of the brain, 3.5 mm below the midcommissural point. The area circled in white indicates the target selected within the GPi by the neurosurgeon. Reproduced with permission from Schaltenbrand G, Bailey P: Introduction to Stereotaxis with an Atlas of the Human Brain. Stuttgart: Thieme, 1959, Vol 2, Plate 56, brain LXXVIII, H.v -3.5. Upper Right: Map of the horizontal section of the brain, 4 mm inferior to the midcommissural point. The area circled in white indicates the surgical target. Reproduced with permission from Talairach J, Tournoux P: Co-Planar Stereotaxic Atlas of the Human Brain. New York: Thieme Medical, 1988, Figure 121. Lower Left: Axial transverse 3D spoiled-gradient MR image obtained under stereotactic conditions. The area circled in black indicates the surgical target. Lower Center: Postoperative control axial transverse MR image demonstrating the position of the electrode under stereotactic conditions (same z coordinate as that shown in upper left panel). Lower Right: Using the 3D spoiled-gradient sequence, the diameter of the center of the artifact (x) created by the presence of the electrode has been positioned on the axial transverse MR image obtained under stereotactic conditions (same z coordinate as that shown in upper left panel). clei that are based on the AC PC line. Consequently, coordinates are usually reported as a range rather than as exact numbers. To compensate for this limited accuracy, personnel at several centers have developed intraoperative clinical and electrophysiological monitoring procedures that can be used with local anesthesia. That approach is uncomfortable for the patient, who is subjected to a long procedure and additional risks related to the multiple cerebral tracts necessary to obtain recordings and ventriculographic images. For the GPi there is no consensus regarding the best targeting method and very little published data documenting the accuracy of any method of localization. 16,17,22 Considering the difference between MR imaging based and atlas-based procedures, we believed that it was necessary to compare both methods. We therefore compared target coordinates within the GPi that were calculated directly on the basis of MR images, with those calculated indirectly on the basis of structures shown in two atlases. Next, we checked whether anatomical data obtained from cadaveric adult brains (in atlases compiled by Schaltenbrand, Talairach, and their colleagues) are applicable to the pediatric population. Clinical Material and Methods Patient Population Between November 1996 and November 2000, 35 patients (17 male and 18 female patients) suffering from drug-resistant generalized dystonia, according to criteria published by Fahn, 9,10 underwent bilateral electrode implantation while in a state of general anesthesia for continuous stimulation of the GPi. The population consisted of 16 adults (mean age 30.6 11.2 years) and 19 children, de- 674

Targeting of the globus pallidus FIG. 2. Upper Left: Reconstructed MR image demonstrating target selection on the brain slice containing the AC. Lower Left: Reconstructed MR image orthogonal to the trajectory, demonstrating the final location of the first contact (1.5 mm under the target). Upper Right: Vertical trajectory superimposed over an MR image. fined in our study as younger than 17 years of age (mean age 11.6 3.5 years). The DYT1 mutation on chromosome 9 was found in nine patients; in 16 the disorder was idiopathic and in 10 the dystonia was secondary (that is, a dystonic disorder that develops as the result of environmental factors that injure the brain). One inclusion criterion was that the GPi boundaries had to be identifiable. Extensive preoperative evaluations, including clinical, neuroradiological, and biochemical investigations, were conducted to identify a specific origin of each patient s dystonia. For assessment of therapeutic efficacy, each patient was evaluated using the BMFDRS 3 both pre- and postoperatively at given intervals; each patient s score was based on a consensus agreement between two physicians (L.C. and P.C.). All patients or their guardians provided written informed consent for the procedure. The protocol was approved by the French National Ethics Committee (Reference Number 98.07.02). After a follow-up period lasting at least 6 months (mean 19.2 10.5 months), the mean improvement in dystonia according to the BMFDRS in these patients was 84.5% for patients with the DYT1 mutation, 80.3% for those with idiopathic dystonia, and 31% for those with secondary dystonia. Targeting Based on MR Imaging The surgical procedure based on MR imaging was developed and validated at our center, and a detailed account of this procedure is published elsewhere. 24 Briefly, in the operating room after general anesthesia had been induced in the patient, the Leksell MR-compatible stereotactic frame was fixed to the patient s head. The patient was then immediately transported to the MR imaging unit. Using 1.5-mm slice thickness, T 1 -weighted MR images were obtained. While the patient was being transported back to the operating room, surgical planning was completed. The neurosurgeon (P.C.) selected the target by using a 3D cursor, following visual recognition of the GPi boundaries on transverse axial sections, without reference to an atlas or to the AC PC line (Fig. 1). The target was chosen vertically on the axial slice at the level of the AC, and horizontally at the junction between the two posterior quarters of the GPi. The x, y, and z coordinates were automatically calculated using dedicated software developed at Val de Grace Hospital, Paris. 7,8 The best electrode trajectory was selected in the direction of the coronal suture and was made as vertical as possible to avoid vessels, ventricles, and sulci. On reconstructed images perpendicular to the electrode trajectory, we checked the position of each contact inside the GPi (Fig. 2). Finally, we checked that this projection crossed the external border of the optic tract. The same procedure was implemented on the other side. Afterward, in the operating room, the stereotactic electrode guiding device was installed and a 14-mm burr hole was made at the level of the predetermined trajectory. No microelectrode recording was used. Electrode implantation was achieved under real-time strict-profile radioscopic control. The distal electrode ends were placed subcutaneously and the patient was transported back to the MR imaging unit, where a control MR image was obtained with the stereotactic frame in place to check the electrode s position, detect any error caused by MR imaging distortions, and detect any hemorrhagic complications. In all patients, the electrodes were checked to ensure that they corresponded with the preoperatively selected target. Comparison Between MR Imaging and Atlas-Based Coordinates: Matrix Transformation. For the purpose of this study, the target coordinates initially obtained on MR images in the Leksell stereotactic reference system (x, y, and z coordinates) were transformed into those used in the AC PC reference coordinates system. 8 This mathematical recalculation required a matrix transformation based on the definition of the AC PC system (three planes) inside the 675

N. Vayssiere, et al. TABLE 1 Distances between target coordinates and the midpoint of the AC PC line* Distance (mm) Coordinate Mean SD Minimum Maximum Median rt hemisphere dx 17.7 1.6 14.8 22.5 17.5 dy 4.6 1.8 0.7 8.2 4.8 dz 1.3 1.7 4.9 2.4 1.3 lt hemisphere dx 17.4 1.9 23.4 13.1 17.3 dy 3.8 2.3 1.3 9.8 3.5 dz 0.9 1.6 3.6 2.7 1.2 L AC PC 22.1 1.7 19.0 25.5 22.1 *Coordinates were calculated on the basis of MR images obtained in 35 patients. Abbreviations: dx, dy, and dz = distance in the x, y, and z directions, respectively; L AC PC = euclidean distance between AC and PC. stereotactic system. To define an appropriate plane, three points are mandatory: 7,8 two points defining the AC PC line, the AC and PC points (x AC, y AC, z AC, x PC, y PC, and z PC ), and a third point, K (x K, y K, and z K ), that is included in an orthogonal plane but not along the AC PC line. The point chosen lay on the falx cerebri. On the basis of the AC, PC, and K coordinates in both the stereotactic reference and the AC PC reference systems (origins and axes of the system), we constructed a transforming matrix, M t, and multiplied it by the target coordinates. We thus obtained the target coordinates in the AC PC reference system, that is, the distances between the target and the midpoint (dx, dy, and dz), or when recentering the system on the midpoint of the AC PC line: dx = x M t, dy = y M t, dz = z M t. Coordinates Based on MR Imaging: Statistical Analysis. A statistical analysis was performed on the coordinates of the 70 targets. The mean and SD, and the minimum and maximum values were calculated and are presented in Table 1. The difference between maximum and minimum coordinates in the three directions ([max(dx) min(dx)], [max (dy) min(dy)], and [max(dz) min(dz)]) reflected the anatomical variability of the target position. The normalized mean and SD, and the minimum and maximum values for the 70 targets are presented in Table 2. They were obtained using the following formulas: Ndx = (L AC PC /Tal AC PC ) dx, Ndy = (L AC PC /Tal AC PC ) dy, and Ndz = (L AC PC / Tal AC PC ) dz, in which Ndx, Ndy, and Ndz represent the normalized values of dx, dy, and dz, respectively; L AC PC is the length of AC PC line in the individual patient under consideration; and Tal AC PC is the length of the AC PC line measured in the atlas of Talairach. 23 TABLE 2 Distances between target coordinates normalized according to length of the AC PC line* Distance (mm) Coordinate Mean SD Minimum Maximum Median rt hemisphere Ndx 18.9 2.2 14.1 23.7 18.6 Ndy 4.8 1.9 0.9 10.1 4.7 Ndz 1.4 1.8 4.9 2.7 1.3 lt hemisphere Ndx 18.7 2.6 23.9 12.5 18.6 Ndy 4.1 2.4 1.3 10.7 3.8 Ndz 0.9 1.7 4.1 2.9 1.3 *Coordinates were calculated on the basis of preoperative MR images obtained in 35 patients. Abbreviations: Ndx, Ndy, and Ndz = normalized distance in the x, y, and z directions, respectively. Targeting Based on the Schaltenbrand Atlas The target was first selected based on the atlas of Schaltenbrand and Wahren. 19 Among the coronal views (called horizontal views) in this atlas, the neurosurgeon chose the plate on which he wanted to position the target for DBS of the GPi in patients with dystonia (Fig. 1). Its projection was then verified on other lateral and frontal views. Once the target had been identified, the distances between the midpoints of the AC and PC, and this target were measured directly on the Schaltenbrand atlas (lateral distance, x Sch = 20 mm; anteroposterior distance, y Sch = 2 mm; and vertical distance, z Sch = 4 mm). Targeting Based on the Talairach Atlas Using the Talairach and Tournoux atlas, 23 the neurosurgeon selected an axial view in which he selected the same target already used in the Schaltenbrand atlas. The coordinates and the distances between the target and the midpoint measured on the Talairach atlas were the following: in the lateral direction, x Tal = 18 mm; in the anteroposterior direction, y Tal = 4 mm; and in the vertical direction, z Tal = 4 mm. Comparison Between MR Imaging and Talairach-Based Coordinates: Normalization Process. The use of the Talairach atlas is based on a homothetic normalization process. A geometric adaptation is applied to dimensions of the patient s brain to be translated into the Talairach atlas system. The normalized data are obtained with the aid of the following formulas: Ndx = (L AC PC /Tal AC PC ) dx, Ndy = (L AC PC /Tal AC PC ) dy, and Ndz = (L AC PC /Tal AC PC ) dz, in which L AC PC is the length of the AC PC line in the patient under consideration and Tal AC PC is the length of the AC PC line measured using the Talairach atlas. Comparison Between MR Imaging and Atlas-Based Coordinates: Statistical Analysis. A statistical comparison was performed between the mean distances to the midpoint of the MR imaging based target (in all three directions) and the same distances deduced from the Schaltenbrand and Talairach atlases, by using a sample-paired Student t-test. Variability of MR Imaging Derived GPi Coordinates With Patient Age and Sex, Surgical Side, and Length of AC PC Line: Statistical Analysis The comparison between mean distances to the midpoint of the MR imaging based target for male and female patients was analyzed using a Student t-test, as was the comparison between the mean distances to the midpoint of MR imaging based target and the length of the AC PC line. The comparison between distances to the midpoint of the MR imaging based target and patient age was analyzed using a nonparametric test of Spearman. One of the authors from the Department of Biomathematics (M.C.P.) performed a descriptive analysis in which 676

Targeting of the globus pallidus the mean SD was used for quantitative variables. Statistical software (nparn way and univariate procedures, SAS version 6.12; SAS Institute, Cary, NC) was used for statistical analysis. Probability values equivalent to or less than 0.05 were considered to be significant. Results Comparison Between MR Imaging Based Coordinates and Schaltenbrand Atlas Based Coordinates In all directions, there were statistically significant differences between coordinates on both left and right sides of the brain (p 0.001 for differences between dx and x Sch, dy and y Sch, and dz and z Sch ). Comparison of MR Imaging Derived Coordinates After Normalization With Talairach Atlas Based Coordinates In the x direction, the difference between Ndx and x Tal was not significant on the left side (p = 0.12), but was statistically significant on the right side (p 0.01). On both left and right sides, in the y and z directions the differences between Ndy and y Tal, and Ndz and z Tal were statistically significant (p 0.01 for both). Coordinates Based on MR Imaging: Statistical Analysis It was possible to identify clearly the GPi boundaries for all candidates for surgery. The mean MR imaging derived coordinates for the 70 studied electrodes were as follows: dx = 17.5 1.8 mm, dy = 4.2 1.9 mm, and dz = 1 1.6 mm. There was a high interindividual anatomical variability of localization according to the midpoint of the AC PC line. The difference between minimum and maximum values of MR imaging derived GPi target coordinates were as follows: [max(dx) min(dx)] = 8.9 mm, [max(dy) min (dy)] = 10 mm, and [max(dz) min(dz)] = 6.8 mm. When applying homothetic normalization of coordinates according to the length of the AC PC line, taking the Talairach-deduced length of the AC PC line as a reference, the difference between minimum and maximum normalized values of GPi target coordinates were as follows: [max(ndx) min(ndx)] = 10.5 mm, [max(ndy) min (Ndy)] = 11.5 mm, and [max(ndz) min(ndz)] = 7.3 mm. Variability in MR Imaging Derived Coordinates Associated With Patient Age In the x direction, no significant difference was found between dx and patient age for either the right or left side (p 0.1). This applied also to the y and z directions for which no significant difference was found between dy and age (p 0.1) or between dz and age (p 0.1) for either the right or left side. The variation in the length of the AC PC line was not statistically different according to the age of patients (p 0.9). Variability in MR Imaging Derived Coordinates Associated With Patient Sex In the x direction, a significant difference was observed for dx in male and female patients (p 0.05): dx was significantly lower in female patients. In the y and z directions, no significant difference was found for dy and dz, respectively, in male and female patients (p 0.9 for both). The length of the AC PC line was not shown to vary significantly with the sex of the patient (p 0.1). Variability in MR Imaging Derived Coordinates With Side of Surgery Values of mean distances in the three directions for right and left hemispheres of the brain are presented in Table 1. No significant differences were found for the values of dx, dy, and dz between right and left sides (p 0.1). Variability in MR Imaging Derived Coordinates With the Length of the AC PC Line In the x direction, no significant difference (p 0.5) was found between dx and the length of the AC PC line in either the left or the right hemisphere of the brain. In the y direction, a significant difference (p 0.005) was found between both variables dy and L AC PC in both left and right hemispheres of the brain. The variation in both variables was in the same direction. In the z direction, no significant difference (p 0.5) was found between dz and L AC PC on either the left or the right sides. Discussion In this study, we sought to compare two methodologies used frequently in stereotactic neurosurgery for targeting the posteroventral portion of the GPi. The first was developed for DBS in children at our center and is based only on 3D MR imaging derived localization of the target. The second, which is more frequently applied at other centers, is essentially based on data synthesized from brain atlases to determine the coordinates of the target used for surgery. In dealing with severely dystonic children, we needed to develop a short procedure (1.5 hours) that could be performed while the patient was in a state of general anesthesia and for which target localization could be based only on 3D MR imaging, without intraoperative microelectrode recording. 4,5,18 Such a procedure, which is based on a strictly defined protocol for the evaluation of distortion before and after surgery in daily practice, has been validated at our center. 24 Indeed, it is now well accepted that such a protocol must be individually tested for accuracy at each medical center, depending on the specifications for the quality of the MR imaging field. In this work, we aimed at an analysis of whether anatomical data obtained from cadaveric brain (using both Schaltenbrand and Talairach atlases) remain relevant in the era of MR imaging, particularly in a pediatric population. Limitations of Atlases of the Human Brain The Schaltenbrand and Wahren atlas 19 is composed of consecutive brain slices (plates) obtained from a limited number of brains. Coordinates calculated on the basis of structures found in this atlas are not obtained from an average brain but actually from a plate (photograph) corresponding to one slice with a given thickness. The number of the brain from which the slice was extracted to be photographed, appears on each plate. Only three brains were used to compile the plates illustrating basal ganglia anato- 677

N. Vayssiere, et al. my. These three brains were obtained from three male cadavers; in two cases the patient was 40 years old at death and in the other the patient was 51 years old. The Talairach atlas is a proportional atlas, but dimensions expressed in millimeters are valid with precision only for a specific brain corresponding to the dimension of the displayed brain (a 60-year-old right-handed European woman). 23 The concept of proportionality is based on an orthogonal grid system. This system is built according to the maximum dimensions of the brain in the three planes of space. This system is supposed to adapt to the dimensions of individual brains. Any measurement calculated on the basis of this proportional atlas must be normalized according to the brain dimensions of the patient to be studied. Furthermore, the whole methodology is based on the assumption that a homothetic anamorphosis between two brains is possible, a concept that has never been proved. 2,6 Correlation of Morphological Characteristics of the GPi With Patient Age and Sex The data contained in the Schaltenbrand and Talairach atlases are based on adult brains. Although in several studies brain structures have been localized according to patient age, 1,11,13 it has not yet been established whether these atlases featuring the adult brain are relevant to the pediatric population. Our data revealed that the position of the pallidum target in reference to an internal landmark (midpoint of the AC PC line) does not depend on patient age, at least after the age of 6 years, which was the age of the youngest girl included in this study. We found a correlation between the target position and patient sex in the lateral direction, dx. This means that in the lateral direction the target used in this study is closer to the midpoint in female than in male patients. We concluded that the sex of the patient must be considered when using data synthesized from atlases. Concept of Target in Stereotactic Neurosurgery Targeting of the GPi based on plates in the Schaltenbrand atlas was extensively described and used by Leksell. 22 He reported on the use of GPi targeting in 38 patients with Parkinson disease in whom stereotactic x-ray filming, and computerized tomography, but not ventriculography, were used. 16,22 He selected the final lesion site, taking into account the patient s reactions to intraoperative electrical stimulation 22 (improvement of tremor, rigidity, and bradykinesia). The coordinates of this target with respect to the midpoint of the AC PC line were x = 18 to 21 mm, y = 2 to 3 mm, and z = 4 to 6 mm. A few years later, Laitinen and colleagues 15 17 reported on the same target, which since has become the reference target for most neurosurgeons involved in pallidal surgery. In the present study, the neurosurgeon (P.C.) targeted the posterior portion of the GPi, as it appears in the atlases, and the coordinates obtained appeared to be closely comparable to those obtained by Leksell. It has long been known that there is substantial individual variation in the coordinates of subcortical nuclei based on the AC PC line. Our study demonstrated that, for a given patient, the GPi target coordinates on MR imaging can be very different from the corresponding coordinates appearing in an atlas (p 0.01). In fact, this result confirms the need for refining the atlas-based localization of the target by performing electrophysiological and intraoperative clinical monitoring while the patient experiences local anesthesia. Based on a few reported studies 12,21 in which MR imaging was used for stereotactic neurosurgery, we can infer that the final target localization was based on the Schaltenbrand atlas. The authors reported a significant difference between data obtained using their indirect form of GPi targeting based on MR imaging and the final position of the electrode following electrophysiological recordings and intraoperative clinical assessment. We now wonder whether the limits of atlas accuracy should not be questioned. In fact, a gold standard still remains to be defined. Based on our results, MR imaging, which provides direct visualization of the nuclei, will probably be a crucial reference tool and allow simplification of the procedure for the patient and surgeon. Anatomical Variability in MR Imaging Derived Coordinates of the GPi Target In our experience, target coordinates used in the treatment of dystonia differ greatly from one patient to another. We attempted to increase the reliability of atlas comparison by homothetic normalization according to the length of the AC PC line. Despite this refinement, the individual anatomy of the basal ganglia cannot be precisely predicted on the basis of data provided in atlases. Length of the AC PC Line and Brain Volume It has long been accepted that the use of atlases for stereotactic neurosurgery is conditioned to data normalization according to specificities of each brain; however, there has been no consensus about the normalization process. Some data have been presented regarding elastic deformation, 2,6,14 normalization based on the length of the AC PC line, 23 or normalization based on the size of the entire brain. 2,6,14,20 The correlation obtained in our study between the GPi target coordinate in the y direction and the length of the AC PC line strengthens the solution of a normalization of this length in the y direction. Nevertheless, there remains no definite solution to this problem. The use of MR imaging based direct targeting, which correlates with the actual position of the structure inside the brain, would remove the need for any kind of normalization. The use of a digital MR imaging atlas (database) could be an interesting solution when theoretical data are mandatory, especially when the target is rendered nonvisible by pathological modifications of the GPi, such as those found in a patient with secondary dystonia. Conclusions We confirm that there is a significant difference from one patient to another between target coordinates obtained using MR imaging. This method provides direct visualization of the target before surgery and of the electrode position thereafter, confirming the accuracy or lack thereof of the overall procedure. Subsequently, the clinical efficacy we obtain is the only criterion for validating the relevance of target selection within the GPi boundaries. We conclude that, as long as high-definition 3D MR imaging provides 678

Targeting of the globus pallidus reliable and direct anatomical visualization of surgical targets, the use of atlases should be undertaken with caution and complete awareness of their limitations. References 1. Autti T, Raininko R, Vanhanen SL, et al: MRI of the normal brain from early childhood to middle age. II. Age dependence of signal intensity changes on T 2 -weighted images. Neuroradiology 36:649 651, 1994 2. Bajcsy R, Lieberson R, Reivich M: A computerized system for the elastic matching of deformed radiographic images to idealized atlas images. J Comput Assist Tomogr 7:618 625, 1983 3. Burke RE, Fahn S, Marsden CD, et al: Validity and reliability of a rating scale for the primary torsion dystonias. Neurology 35: 73 77, 1985 4. Coubes P, Echenne B, Roubertie A, et al: Traitement de la dystonie generalisee a debut precoce par stimulation chronique bilaterale des globus pallidus internes. A propos d un cas. Neurochirurgie 45:139 144, 1999 5. Coubes P, Roubertie A, Vayssiere N, et al: Treatment of DYT1- generalised dystonia by stimulation of the internal globus pallidus. Lancet 355:2220 2221, 2000 (Letter) 6. Davatzikos C: Spatial normalization of 3D brain images using deformable models. J Comput Assist Tomogr 20:656 665, 1996 7. Derosier C, Buee C, Ledour O, et al: MRI and stereotaxis. Choice of an approach route on an independent console. J Neuroradiol 18:333 339, 1991 8. Dormont D, Cornu P, Pidoux B, et al: Chronic thalamic stimulation with three-dimensional MR stereotactic guidance. AJNR 18:1093 1107, 1997 9. Fahn S: Concept and classification of dystonia. Adv Neurol 50: 1 8, 1988 10. Fahn S: The varied clinical expression of dystonia. Neurol Clin 2:541 554, 1984 11. Giedd JN, Snell JW, Lange N, et al: Quantitative magnetic resonance imaging of human brain development: ages 4-18. Cereb Cortex 6:551 560, 1996 12. Guridi J, Gorospe A, Ramos E, et al: Stereotactic targeting of the globus pallidus internus in Parkinson s disease: imaging versus electrophysiological mapping. Neurosurgery 45:278 289, 1999 13. Holland BA, Haas DK, Norman D, et al: MRI of normal brain maturation. AJNR 7:201 208, 1986 14. Kochunov PV, Lancaster JL, Fox PT: Accurate high-speed spatial normalization using an octree method. Neuroimage 10: 724 737, 1999 15. Laitinen L, Johansson GG: Stereotaktisk behandling av hyperkinesier. Nord Med 75:676 679, 1966 16. Laitinen LV: Brain targets in surgery for Parkinson s disease. Results of a survey of neurosurgeons. J Neurosurg 62: 349 351, 1985 17. Laitinen LV, Bergenheim AT, Hariz MI: Ventroposterolateral pallidotomy can abolish all parkinsonian symptoms. Stereotact Funct Neurosurg 58:14 21, 1992 18. Roubertie A, Echenne B, Cif L, et al: Treatment of early-onset dystonia: update and a new perspective. Childs Nerv Syst 16: 334 340, 2000 19. Schaltenbrand G, Wahren W: Atlas for Stereotaxy of the Human Brain, ed 2. Stuttgart: Georg Thieme, 1977 20. Sorlie C, Bertrand O, Yvert B, et al: Matching of digitised brain atlas to magnetic resonance images. Med Biol Eng Comput 35: 239 245, 1997 21. Starr PA, Subramanian T, Bakay RA, et al: Electrophysiological localization of the substantia nigra in the parkinsonian nonhuman primate. J Neurosurg 93:704 710, 2000 22. Svennilson E, Torvik A, Lowe R, et al: Treatment of Parkinsonism by stereotactic thermolesions in the pallidal region. A clinical evaluation of 81 cases. Acta Psychiatr Neurol Scand 35:358 377, 1960 23. Talairach J, Tournoux P: Co-Planar Stereotaxic Atlas of the Human Brain: 3-Dimensional Proportional System: An Approach to Cerebral Imaging. New York: Thieme Medical, 1988 24. Vayssiere N, Hemm S, Zanca M, et al: Magnetic resonance imaging stereotactic target localization for deep brain stimulation in dystonic children. J Neurosurg 93:784 790, 2000 Manuscript received May 7, 2001. Accepted in final form December 3, 2001. Address reprint requests to: Philippe Coubes, Unité de Recherche sur les Mouvements Anormaux de l Enfant, Département de Neurochirurgie B, Centre Hospitalier Universitaire Gui de Chauliac, 80 Avenue Augustin Fliche, 34295 Montpellier Cedex 5, France. email: p-coubes@chu-montpellier.fr. 679