Stereotactic investigation of limbic epilepsy using a multimodal image analysis system

Transcription

1 J Neurosurg 73: , 1990 Stereotactic investigation of limbic epilepsy using a multimodal image analysis system Technical note MICHEL F. LI~VESQUE, M.D., JINGXI ZHANG, PH.D., CHARLES L. WILSON, PH.D., ERIC J. BEHNKE, B.S., RONALD n. HARPER, PH.D., ROBERT B. LUrKIN, M.D., JEROME ENGEL, JR., M.D., PH.D., AND PAUL H. CRANDALL, M.D. Division of Neurosurgery and the Departments of Neurology, Anatomy and Cell Biology, Radiological Sciences, and the Brain Research Institute, University of California School of Medicine, Los Angeles, California ~" A methodology has been developed for stereotactic investigation of limbic epilepsy using an image-analysis system that simultaneously displays different structural and functional images of the brain. The validity and accuracy of this system were established with phantom studies. Surgical planning and electrode implantation are guided by stereotactic magnetic resonance imaging, digital subtraction angiography, and positron emission tomography. This methodology provides the spatiotemporal relationship of cerebral structure and function necessary to identify seizure onset and propagation in human limbic system epilepsy. KEY WORDS ~ stereotactic surgery 9 limbic system 9 epilepsy 9 neuroimaging 9 stereoelectroencephalography 9 image analysis system I NVASIVE recording of cerebral structures is sometimes necessary to establish the epileptogenic area of partial epilepsies amenable to surgical therapy.8.12 Depth electrodes implanted for prolonged recording of stereoelectroencephalography (SEEG) provide the accuracy and spatiotemporal resolution necessary to demonstrate seizure onset and propagation paths in patients with nonlocalized limbic system epilepsy.~'6 Previously, electrode implantation at our center incorporated the Talairach system, modified for symmetrical bitemporal sampling of superficial neocortical and deep mesiolimbic structures. 2'8 This technique assumed a symmetrical brain and relied on intracerebral landmarks upon which indirect visualization of structures was based using the Talairach stereotactic atlases. ~3'~4 Since the validity of recorded events during SEEG depends on the accuracy of electrode placement, we modified this approach to combine magnetic resonance (MR) imaging, digitized subtraction angiography (DSA), and computerized tomography (CT) in order to provide direct vascular and structural visualization. Stereotactic positron emission tomography (PET) was also integrated with the system, since PET studies reveal regions of cerebral dysfunction in partial epilepsy that differ spatially from electrophysiological abnormalities or pathological substrates? Materials and Methods Stereotactic Frame and Fiducial Markers The stereotactic apparatus is a modified Leksell frame named "OBT."~ This device is compatible with CT, MR imaging, DSA, and (after modifications made at the University of California, Los Angeles) with PET. Its construction minimizes artifact in these imaging modalities and the system uses a cartesian coordinate system with the x axis origin located at the most anterior point of the sagittal plane and the y axis origin at the inferior portion of the frame in the coronal axis, while the z axis extends positively from the midline to the right or negatively to the left along the axial plane. Four sets of Plexiglas plates provide fiducial markers on each side and on top of the stereotactic frame. Three sets contain a Z-shaped channel filled with an appropriate contrast material for each image modality and are * Modified Leksell frame manufactured by Tipal Instruments, Montreal, Quebec, Canada. 792 J. Neurosurg. / Volume 73/November, 1990

2 Stereotactic investigation of limbic epilepsy FIG. 1. Hexagonal and grid phantom studies, a: Grid phantom used for magnetic resonance (MR) distortion studies in the coronal plane within the frame, b: The hexagonal frame during stereotactic MR and computerized tomography studies displayed in two windows, c: Digital fluoroscopy image of the grid phantom. The fiducial marks are four points on each plate placed anteriorly and posteriorly, d: Stereotactic MR image of the grid phantom showing minimal distortion created by the frame. temporarily attached to the stereotactic frame. Aluminum tubing is using during CT scanning, copper sulfate solution (7 gin/liter) is used for MR images, and the channels are filled with a gallium emitter for PET scans. The plane of section is calculated from the location of the center arm of the Z-shaped marker in relation to the two parallel end bars. For DSA, the fiducial markers consist of four 1-mm stainless steel disks placed equidistantly at the four quadrants of the Plexiglas plates located on either side of the head for lateral views, or at the front and back of the frame for anteroposterior views. Markers closest to the x-ray source will form a larger rectangle on the x-ray film because of beam divergence, and thus differentiate the side injected and provide data for computer analysis of depth of field. After informed consent is obtained, the OBT stereotactic frame is placed on the patient's head using local anesthesia supplemented with short-acting neuroleptic agents. Stereotactic MR, PET, DSA, and CT studies are then performed, following which the frame is removed at the bedside. The device is replaced under general anesthesia for the surgical procedure once the stereotactic images are analyzed and targets are selected. Localizing tings placed on each of the pins that penetrate the outer table of the skull insure proper repositioning of the frame; the position is also documented by a skull x-ray film. Imaging Studies Stereotactic MR Imaging. Sagittal, coronal, and axial images are obtained on a 0.3-tesla unit.? We use t Magnetic resonance imaging unit, Model B-3000, manufactured by FONAR Corp., Melville, New York. a customized surface coil that fits closely around the frame to increase the signal-to-noise ratio. Inversionrecovery sequences (IR 1276 msee, TR 300 msec/te 30 msec) allow accurate definition of the cerebrospinal fluid-brain and gray-white matter interface. Slice thickness is 4.9 mm, slice interval is 5.1 ram, and a resolution is 512 levels by 256 fast-fourier transformation columns. Three excitations are used in the coronal plane and two in the axial and sagittal planes. All imaging is performed with the stereotactic frame anchored to a custom-made frame-adaptor placed over the sliding table of the MR imager, allowing exact reproducibility between preoperative and electrode verification imaging studies. Stereotactic DSA. Digital angiograms in both anteroposterior and lateral projections are obtained with a digital radiography systems using a standard femoral catheterization approach. Images of arterial, capillary, and venous phases are selected for further analysis. Stereoscopic studies with a 5* projection angle may also be performed. Stereotactic PET For PET studies,w a stereotactic frame adaptor is fixed to the tomograph sliding table. The patient is injected with 5 mci of 18-fluorodeoxyglucose, and 15 simultaneous axial planes with a centerto-center interslice distance of 6.75 mm are obtained 40 minutes later. Images are reconstructed by filtered backprojection to an image resolution dependent on the field of view and a resolution of 128 x 128 pixels. Stereotactic CT. Axial scans are obtained using 10 x 10 mm sections.l[ These studies serve as an internal test to assure accuracy of other modalities. Phantom Studies. Preliminary studies using a cylindrical phantom filled with copper sulfate solution were made in the stereotactic frame, and MR sections were obtained to assess constant peripheral distortion. A grid containing hollow tubing at a fixed distance within the frame was used as a second phantom for the MR image and digital angiogram. The grid was fixed to the stereotactic frame, and sequential images were obtained following injection of contrast material (copper sulfate) into the hollow tubing. Linearity was assessed in the stereotactic frame. Finally, attenuation measurements for the frame were made in the PET scanner with the established phantom. Description of the System The system consists of a VAXStation II/GPX computer,* equipped with 16 megabytes of memory and a Digital radiography system manufactured by ADAC, San Jose, California. w Siemens-CTI 831 tomograph, manufactured by Siemens, Knoxville, Tennessee. 1[ GE 8800 unit manufactured by General Electric, Milwaukee, Wisconsin. * VAXStation II/GPX computer manufactured by Digital Equipment Corp., Maynard, Massachusetts. J. Neurosurg. / Volume 73/November,

3 M. F. Lrvesque, et al. FIG. 2. Multiwindow display of magnetic resonance studies used for surgical planning. Temporal lobe targets are selected on the sagittal, coronal, and axial views. The system of dynamic cursors interrelates different windows; the "master" cursor is shown in red on the right amygdala, and "slave" cursors are seen in white on the other planes. Serial projections in each window allow visualization of electrode trajectories. Targets are labeled, and their x, y, and z coordinates are displayed. 320-megabyte hard disk. Data are transferred from MR, CT, DSA, and PET studies using either TK-50 or ninetrack magnetic tape. The display terminal is a 19-in. RGB 8-bit plane monitor displaying pixels. A VAX VMS common file structure is created after the MR, CT, DSA, and PET study data are converted for entry; the file includes equipment parameters, image size, orientation, slice thickness and intervals, image side, field of view, and patient identification. The image analysis program uses a mouse-driven pull-down menu system, allowing easy user interface.~5 Selected images are retrieved from the hard disk memory and are simultaneously displayed in separate windows. Different planes from a single modality or multimodal images can be displayed and analyzed at the same time. A system of cursors interconnects each stereotactic image, and x, y, and z coordinates of the point of interest are dynamically displayed. Labeled points can be transferred from one image window to another independently of the plane of section. Regions 794 of interest and anatomical boundaries of brain structures can be drawn and correlated across modalities with high accuracy. The system also provides image processing functions such as image subtraction, gray scale correction, image zooming, and pseudocolor. Results Phantom Studies Accuracy and distortion of the MR images in the axial, coronal, and sagittal planes were determined with the cylindrical and grid phantoms (Fig. 1). The onplane geometric distortion was 1 m m in any plane, and an excellent correlation was demonstrated between the computerized coordinate system and phantom targets. The grid phantom was also used in anteroposterior, lateral, and stereoscopic angiograms. Pin-cushion distortion was detected, mainly at the periphery where it reached 2 mm. J. Neurosurg. / Volume 73/November, 1990

4 Stereotactic investigation of limbic epilepsy F~G. 3. Multimodal display of a right sagittal magnetie resonance image (lower left) and fight internal carotid angiograms at the arterial (upper left), capillary (upper right), and venous (lower right) phases. Trajectories are corrected to avoid cerebral vessels during implantation. Multimodal Stereotactic Imaging System Figure 2 presents the multiwindow MR imaging display in the sagittal, coronal, and axial planes. A system of dynamic cursors integrates all displayed images and consists of two cross-hairs that overlap when identical points are reached on different planes or different modalities. Parallel analysis of the same target in all three planes is made in the other display windows by a yoked set of cursors. The targets can rapidly be plotted with the use of templates of stereotactic atlases that are then modified according to anatomical variations in each brain, and corrected for brain size or inherent pathology. Standardized placement of electrodes within temporal lobe structures includes sampling of neocortex, amygdala, pes hippocampus, and hippocampal gyrus. Extratemporal sites are selected according to symptomatology, extracranial electrophysiological data, and regional hypometabolism as seen on PET. All selected targets can be automatically transferred from one plane or slice to another or to another imaging modality; for example, from MR imaging to DSA or from PET to J. Neurosurg. / Volume 73/November, 1990 MR imaging. The mouse-driven cursor can select any point in space and display its x, y, and z coordinates. This point can be labeled, stored, and retrieved during the display of any further studies. Different phases of the angiogram are seen in Fig. 3 in parallel to sagittal MR images. Entry points and trajectories are corrected to avoid vascular structures as observed during early arterial to late venous phases, Stereoscopic studies were performed initially to differentiate surface vessels from deeper sulcal arteries, and indirectly to determine gyri and sulci anatomy; however, surface anatomy was clearly better defined using sagittal MR images. Stereotactic PET and MR studies are displayed in parallel at the same axial section in Fig. 4. A large hypometabolic area is seen over the mesial and lateral temporal lobe on the stereotactic PET scan which was sampled with SEEG. The corresponding stereotactic MR axial slice shows a region of interest drawn on anatomical landmarks which is automatically transposed on the stereotactic PET. 795

5 M. F. Lrvesque, et al FIG. 4. Left: Stereotactic structure-function correlation with regions of interest drawn on anatomical landmarks obtained with an axial magnetic resonance image (slice thickness: 5.3 mm). Right: The regions of interest are transposed to the corresponding axial slice of the stereotactic positron emission tomography scan (slice thickness: 6.75 mm). There is a large area of relative cerebral hypometabolism over the entire left temporal lobe, which was sampled with stereoelectroencephalography. Discussion Phantom Studies Knowledge of field inhomogeneities 7 and correction of spatial distortion ~' are essential elements for accurate MR stereotactic imaging. Accuracy and linearity of this system, as demonstrated by phantom studies, show a maximum distortion of 1 m m on sagittal MR images at the field periphery. Other centers have reported greater distortion using a 1.5-tesla magnet; our 0.3-tesla unit produces satisfactory resolution of targeted structures but cannot display higher definition of the amygdala or hippocampus. Our phantom studies have also shown that angiographic accuracy with DSA is not as precise as with MR guidance because of the pin-cushion distortion common to such studies. Multimodal Stereotactic Image Analysis The multimodal stereotactic image analysis (MUSIS) systemt has a number of important advantages over previous stereotactic methodologies. The ability to simultaneously display and analyze several different planes of MR imaging in a stereotactic environment provides a greatly improved understanding of the spat Multimodal stereotactic image analysis system software (MUSIS) is patented and available through the UCLA Office of Contract and Grant Administration, c/o Wade A. Bunting, 1400 Ueberroth Building, Le Conte Avenue, Los Angeles, California tial organization of cerebral structures that was not obtainable using ventriculography and stereotactic atlases alone or with stereoscopic angiography. Surgical planning of trajectories and target selection for electrode implantation are made directly without correction for magnification. Furthermore, the cursor system that interconnects different images selected on the display terminal increases the accuracy of implantation. Interindividual and interhemispheric variability of cerebral structures, atrophy, or distortion are directly visualized. Templates of different lines of reference taken from stereotactic atlases can be projected, mainly the linea temporalis and the anterior-posterior commissure (ACPC) lines. Templates assist in the selection of targets within ill-defined telencephalic anatomical regions such as the supplementary motor area. Attempts to integrate computerized imaging into the methodology of functional stereotactic procedures were pioneered at the Montreal Neurological Institute (MNI) 9 and the Mayo Clinic) The MNI system uses the corpus callosum to transpose stereotactic DSA information to MR and PET studies performed in nonstereotactic conditions, similar to the Talairach approach using the AC-PC line as a reference to the proportionate grid and stereotactic atlas. However, the corpus callosum is not an ideal anatomical reference because it also assumes complete symmetry of the brain and it is indirectly visualized with the angiogram, thus introducing potential errors during target selection. J. Neurosurg. / Volume 73/November, 1990

6 Stereotactic investigation of limbic epilepsy None of these systems use stereotactic PET in the presurgical planning of the procedure. Direct stereotactic MR guidance with inversion-recovery sequences clearly delineates gray-white matter and brain-csf boundaries. The use of multiple planes allows cross-correlation of localization and increases the accuracy of coordinate selection. Intraoperative application and reproducibilty of imaging studies are major advantages of this stereotactic technique. The immediate application of integrating stereotactic PET into our surgical planning is demonstrated by targeting hypometabolic areas and recording from them with SEEG. A more precise correlation can be obtained with both electrophysiological events and video-behavior analysis during telemetry, and this functional targeting may improve localization of the epileptic focus. This information may become complementary to structural information and hence be closer to Horsley and Clarke's initial concept ofstereotactic surgery, 4 defining both cerebral structure and function. Acknowledgments The authors thank Drs. John Mazziotta and Michael Phelps for their advice regarding integration of stereotactic PET images and Dr. Donald P. Becker for his support in the development of this system. References 1. Bancaud J, Talairach J, Bonis A, et al: La St6r6o-Electro- Enc6phalographie dons repilepsie. Paris: Masson, Crandall PH: Developments in direct recordings from epileptogenic regions in the surgical treatment of partial epilepsies, in Brazier MAB (ed): Epilepsy. Its Phenomena in Man. New York: Academic Press, 1973, pp t0 3. Engel J Jr, Brown W J, Kuhl DE, et al: Pathological findings underlying focal temporal lobe hypometabolism in partial epilepsy. Ann Neurol 12: , Horsley V, Clarke RH: The structure and functions of the cerebellum examined by a new method. Brain 31: , Kelly PJ, Sharbrough FW, Kall BA, et al: Magnetic resonance imaging-based computer assisted stereotactic resec- tion of the hippocampus and amygdala in patients with temporal lobe epilepsy. Mayo Cliu Proc 62: , Munari C, Bancaud J: The role of stereo-electroencephalography (SEEG) in the evaluation of partial epileptic seizures, in Porter R J, Morselli PL (eds): The Epilepsies. London: Butterworths, 1985, pp O'Donnell M, Edelstein WA: NMR imaging in the presence of magnetic fields in homogeneities and gradient field non-linearities. Med Phys 12:20-26, Ojemann GA, Engel J Jr: Acute and chronic intracranial recording and stimulation, in Enget J Jr (ed): Surgical Treatment of the Epilepsies. New York: Raven Press, 1987, pp Olivier A, Peters TM, Clark JE, et al: Int6gration de rangiographie num6rique, de la r6sonance magn6tique, de la tomodensitom&rie et de la tomographie par 6mission de positrons en st6r6otaxie. Rev EEG Neurophysiol Clin 17:25-43, 1987 I0. Peters TM, Clark JE, Olivier A, et al: Integrated stereotaxic imaging with CT, MR imaging and digital subtraction angiography. Radiology 161: , Schad L, Lott S, Schmidt F, et al: Correction of spatial distortion in MR imaging: a prerequisite for accurate stereotaxy. J Compnt Assist Tomogr 11: , Talairach J, Bancaud J, Szikla G, et al: Approche nouvelle de la neurochirurige de l'6pilepsie. M6thodologie st6r6otaxique et r6sultats th6rapeutiques. Nenroehirurgie 20 (Snppl 1):1-240, Talairach J, David M, Tournoux P, et al: L'Exploratiou Chirurgicale St6r6otaxique du Lobe Temporal dons repilepsie Temporale: Rep6rage Anatomique St6r~otaxique et Technique Chirurgicale. Paris: Masson, Talairach J, Szikla G, Tournoux P, et al: Atlas d'anatomie St6r6otaxique dn T61enc6phale. Paris: Masson, Zhang J, L6vesque MF, Wilson CL, et al: Multimodality imaging for stereotactic surgery. Radiology 175: , 1990 Manuscript received September 21, Accepted in final form May 11, The University of California is the sole proprietor of the MUSIS software, and none of the authors has any proprietary interest in it. Address reprint requests to: Michel F. L6vesque, M.D., F.R.C.S.(C), Division of Neurosurgery, UCLA School of Medicine, 760 Westwood Plaza, Los Angeles, California J. Neurosurg. / Volume 73/November,