Neurosurgery 56[ONS Suppl 1]:ONS-98 ONS-109, techniques such as magnetoencephalography

Transcription

1 TECHNIQUES AND APPLICATIONS Kyousuke Kamada, M.D. Department of Neurosurgery, University of Tokyo, Tokyo, Japan Yutaka Sawamura, M.D. Department of Neurosurgery, Hokkaido University, Sapporo, Japan Fumiya Takeuchi, Ph.D. Research Institute for Electronic Science, Hokkaido University, Sapporo, Japan Hideaki Kawaguchi, M.D. Department of Clinical Medicine, Hokkaido University, Sapporo, Japan Shinya Kuriki, Ph.D. Research Institute for Electronic Science, Hokkaido University, Sapporo, Japan Tomoki Todo, M.D. Department of Neurosurgery, University of Tokyo, Tokyo, Japan Akio Morita, M.D. Department of Neurosurgery, University of Tokyo, Tokyo, Japan FUNCTIONAL IDENTIFICATION OF THE PRIMARY MOTOR AREA BY CORTICOSPINAL TRACTOGRAPHY OBJECTIVE: For quick and stable identification of the primary motor area (PMA), diffusion tensor imaging (DTI) data were acquired and corticospinal tractography was mathematically visualized. METHODS: Data sets of DTI, anatomic magnetic resonance imaging, and functional magnetic resonance imaging with finger-tapping tasks were acquired during the same investigation in 30 patients with a brain lesion affecting the motor system. Off-line processing of DTI data was performed to visualize the corticospinal tract, placing a seed area in the cerebral peduncle of the midbrain, where the corticospinal tract is densely concentrated. Somatosensory evoked magnetic fields and intraoperative cortical somatosensory evoked potentials were recorded with electrical stimulation of the median nerve to confirm the results of the corticospinal tractography. RESULTS: Functional magnetic resonance imaging and somatosensory evoked magnetic fields failed to identify the PMA in eight patients (16.7%) and one patient (3.8%) investigated, respectively, because of cortical dysfunctions caused by brain lesions. DTI data were acquired within 3 minutes without patient tasks. Using the appropriate seed area and fractional anisotropy, corticospinal tractography successfully indicated the PMA location in all patients. The suspected PMA and central sulcus locations were confirmed by the cortical somatosensory evoked potentials. CONCLUSION: Corticospinal tractography enables identification of the PMA and is beneficial, particularly for patients who present with dysfunction of the PMA. KEY WORDS: Corticospinal tract, Diffusion tensor imaging, Magnetoencephalography, Primary motor area, Tractography Neurosurgery 56[ONS Suppl 1]:ONS-98 ONS-109, 2005 DOI: /01.NEU EF Yoshitaka Masutani, Ph.D. Department of Radiology, University of Tokyo, Tokyo, Japan Shigeki Aoki, M.D. Department of Radiology, University of Tokyo, Tokyo, Japan Takaaki Kirino, M.D. Department of Neurosurgery, University of Tokyo, Tokyo, Japan Reprint requests: Kyousuke Kamada, M.D., Department of Neurosurgery, Faculty of Medicine, University of Tokyo, Bunkyo-ku, Hongo 7-3-1, Tokyo , Japan. kamady-k@umin.ac.jp Received, November 12, Accepted, April 8, Noninvasive functional brain mapping techniques such as magnetoencephalography (MEG) and functional magnetic resonance imaging (fmri) have been used routinely to identify the sensorimotor area for preoperative surgical planning. MEG reflects the intracellular electrical current flow in the brain, providing direct information regarding neural activity. The somatosensory evoked magnetic field (SEF) has become the gold standard for identification of the central sulcus (CS) (5, 10, 11). Reduced SEF amplitudes, however, may occasionally cause unacceptable errors of dipole localization in patients with severe sensory impairment (8). It is well known that neuronal activation induces an increase in local blood flow, and fmri detects local changes of the magnetic susceptibility of blood, demonstrating the blood oxygen level-dependent contrast. The crucial limitations encountered in fmri-based brain mapping are the various degrees of venous drainage architectures and regional hemodynamic responses in pathological brain conditions (1, 6, 8, 19). Furthermore, patients with cortical dysfunction such as hemiparesis or dementia can rarely achieve self-paced finger tapping, which is a typical fmri task for CS identification. There still remain several technical issues with noninvasive brain mapping, and the results may be significantly affected by the patient s condition. It is thus important to find a new technique for this purpose that can quickly complete the data acquisition process and subjectively make a secure identification of the sensorimotor area. Diffusion-weighted imaging has been demonstrated to be directionally dependent on ONS-98 VOLUME 56 OPERATIVE NEUROSURGERY 1 JANUARY

2 CORTICAL MAPPING BY TRACTOGRAPHY water molecule diffusion in the white matter (anisotropic diffusion), which has been attributed to the organization of axonal fibers and their myelin sheaths (3). Anisotropy is quantified with a full description of diffusion in the region of interest (the diffusion tensor) by measuring changes in the nuclear magnetic resonance signal with diffusion sensitization along at least six non-colinear directions. The most important index for anisotropy among the several derived indices is fractional anisotropy (FA). It has recently become possible to visualize the major axonal fascicles in vivo in the human brain by selecting regions with high FA values as seed areas for the three-dimensional tracking process (tractography) (15, 16, 20, 21). Recent studies have demonstrated the major fascicles running in the craniocaudal direction as the corticospinal tract (CST) and stressed its clinical usefulness for presurgical planning and functional prediction in cases of lesions adjacent to the CST (12, 16). The fascicles along the craniocaudal direction theoretically contain not only the CST (the motor tracts) but the somatosensory tracts, depending on locations of seed areas. It is thus important to demonstrate the target fascicles such as the CST with minimum contamination of other tracts on tractography. The brainstem is a region characterized by densely packed fibers traveling to and from the cerebrum and cerebellum (14, 18). The CST and the somatosensory tracts are separately located in the cerebral peduncle and medial lemniscus (ML) in the midbrain, respectively. Furthermore, it is known that the CST is anatomically composed of more than 10 6 fibers arising from the primary motor area (PMA, Brodman s area 4) rather than the descending fibers from the somatosensory cortex (4, 7). On the basis of this knowledge, we expected that corticospinal tractography would become one of clinical tools to indicate the PMA location using appropriate seed areas and FA values. The expected PMA locations on tractography were compared with those identified by fmri, MEG, and the cortical somatosensory evoked potentials (SEPs). In this study, we specially focused on identifying the PMA using corticospinal tractography and confirmed its reliability by other mapping techniques. PATIENTS AND METHODS Patients Studies were performed on 30 patients with a mass lesion affecting the primary motor cortex, including astrocytic tumor (n 21), metastatic brain tumor (n 1), arteriovenous malformation (n 2), cavernous angioma (n 2), and meningioma (n 4). Because the cortical SEPs were used to confirm the CS location during surgery, patients who did not undergo cortical SEP testing were excluded from this analysis. The demographic data of all patients are summarized in Table 1. This project was approved by the ethics committees of our institutions, and written informed consent was obtained from each subject or his or her family before participation in the studies. Magnetic Resonance Protocols All magnetic resonance imaging (MRI) studies were performed as a single MRI investigation on 1.5-T whole-body MRI scanners with echo-planar capabilities and a standard whole-head transmit-receiver coil (Magnetom VISION, Siemens AG, Erlangen, Germany; and TwinSpeed, General Electric Medical Systems, Milwaukee, WI). Diffusion Tensor Imaging We used a single-shot, spin echo, echo-planar sequence with a TR of 5000 to 6000 milliseconds and a TE of 85 to 95 milliseconds, acquiring 26 interleaved, contiguous, 4-mm axial images with no cardiac triggering. A data matrix of over a field of view (FOV) of mm was obtained, acquiring 128 echoes per excitation. Diffusion weighting was performed along six independent axes, using b values of 0 and 1000 s/mm 2. A single echo-planar imaging set took 25 seconds and was repeated six times to increase the signal-to-noise ratio. Realignment of these seven sets of images and compensation for eddy current-induced morphing were performed on an equipped workstation with one of the MRI scanners. Threedimensional anatomic MRI data of each patient s head were obtained, consisting of 96 sequential 1.8-mm-thick axial slices with a resolution of pixels over a FOV of 260 to 280 mm with a turbo fast low-angle shot (FLASH) or spoiled gradient-recalled acquisition in steady-state sequence. Because of the different head size and position of each subject, we selected a large FOV (range, mm) that could contain the entire head, fixing the same center of the FOV on the x and y axes through all image sessions. This procedure enabled us to perform the simple image registration in different image sessions. During the whole MRI investigation, patients kept their eyes closed and foam cushions were used to immobilize the head. fmri fmri was performed with a T2*-weighted echo-planar imaging sequence (TE, 62 ms; TR, 114 ms; flip angle, 90 degrees; slice thickness, 5 mm; slice gap, 2.5 mm; FOV, 280 mm; matrix, ; 10 slices), resulting in an acquisition time of 2 seconds for each fmri volume. Each fmri session consisted of three dummy scan volumes and three activation and four baseline (rest) periods. During each period, 5 image volumes were collected, yielding a total of 38 imaging volumes. fmri data of the motor evoked response were acquired by selfpaced finger tapping at a constant rhythm of approximately one cycle per second. After the data acquisition, a motion detection program (MEDx; Medical Numerics, Sterling, VA) estimated motion artifact through each fmri session. fmri sessions containing motion artifact of more than 25% of the pixel size were discarded. After omitting the first three dummy volumes and NEUROSURGERY VOLUME 56 OPERATIVE NEUROSURGERY 1 JANUARY 2005 ONS-99

3 KAMADA ET AL. TABLE 1. Demographic data of 30 patients with lesions affecting the central motor system a Patient no. Age (yr)/sex Lesion location Symptoms Histological findings Outcome of surgery 1 15/F Rt frontal No symptom Astrocytoma (Grade II) No deficit 2 37/F Lt frontal No symptom Astrocytoma (Grade II) No deficit 3 39/F Lt frontal Severe rt hemiparesis Astrocytoma (Grade II) No deterioration 4 37/M Lt frontal Mild rt hemiparesis Astrocytoma (Grade II) No deterioration 5 34/M Rt frontal No symptom Astrocytoma (Grade II) No deficit 6 28/F Lt frontal Generalized seizure Astrocytoma (Grade II) Improved 7 63/M Rt frontal No symptom Astrocytoma (Grade II) No deficit 8 45/M Rt frontal No symptom Astrocytoma (Grade III) No deficit 9 40/F Lt frontal Generalized seizure (claustrophobia) Astrocytoma (Grade III) Transient facial palsy 10 32/M Lt frontal Mild rt hemiparesis Astrocytoma (Grade III) No deficit 11 43/M Lt frontal Generalized seizure Astrocytoma (Grade III) No deficit 12 32/M Lt frontal Mild rt hemiparesis Astrocytoma (Grade III) No deficit 13 32/M Lt frontal No symptom Astrocytoma (Grade III) Transient rt hemiparesis 14 53/F Rt frontal Severe rt hemiparesis Glioblastoma No deterioration 15 38/M Rt frontal Mild lt hemiparesis Glioblastoma Improved 16 63/M Lt frontal Generalized seizure Glioblastoma Transient aphasia 17 52/F Lt frontal Moderate rt hemiparesis Glioblastoma No deterioration 18 49/F Rt frontal Severe lt hemiparesis Glioblastoma Improved 19 52/M Lt frontal Mild rt hemiparesis Glioblastoma No deterioration 20 49/M Lt frontotemporal Mild rt hemiparesis and aphasia Glioblastoma Rt hemiparesis 21 61/M Lt frontal Mild rt hemiparesis and dysarthria Metastasis No deterioration 22 24/M Rt frontal Mild lt hemiparesis AVM No deterioration 23 64/M Lt frontal Mild rt hemiparesis AVM No deterioration 24 50/M Rt parietal Mild lt hemiparesis Cavernous angioma Transient hemiparesis 25 55/M Rt frontal Generalized seizure Cavernous angioma No deficit 26 16/F Lt frontal Mild rt hemiparesis and dysarthria Meningioma Improved 27 65/M Lt parietal Moderate rt hemiparesis and lt hemisensory deficit Meningioma Improved 28 58/F Lt parietal No symptom Meningioma No deficit 29 56/F Lt frontoparietal Mild rt hemiparesis Meningioma Improved 30 74/F Lt parieto-occipital Mild confusion Glioblastoma No deterioration a Rt, right; Lt, left; AVM, arteriovenous malformation. ONS-100 VOLUME 56 OPERATIVE NEUROSURGERY 1 JANUARY

4 CORTICAL MAPPING BY TRACTOGRAPHY applying a Gaussian spatial filter (7 mm in half-width) for each session, functional activation maps were calculated with a cross-correlation analysis between the measured and expected activation time courses for each voxel using Dr. View (Asahi Kasei Medical Co., Tokyo, Japan). Pixels with a z score of more than 1.0 were accepted. Image Registration Because all MRI scans were obtained during the same MRI investigation using a similar FOV of 260 to 280 mm with the same image center on x and y axes, integration of the diffusion tensor imaging (DTI), fmri, and anatomic MRI data was simply accomplished by adjusting the z axis center of all sessions. After the rough registration, we proceeded to refine registration using the automatic multimodality image registration algorithm provided with Dr. View (2). This algorithm made several clusters on the basis of the anatomic MRI using a k-means algorithm and applied the created clusters to functional data sets. Within each defined cluster on the functional data sets, the algorithm iteratively updated the registration parameters using affine transformation to minimize results of the cost function, which evaluated the spatial difference between anatomic and functional data sets (2). After the registration, functional data (7 sets of 26 diffusion tensor slices and 38 sets of 10 fmri slices) were interpolated and resliced on the basis of the anatomic MRI. Tensor Calculation and Tractography The diffusion tensor at each pixel of the registered DTI data was calculated, and three-dimensional fiber tracking was then performed using our own software ( Volume-one and dt1 ; Interpolation along the z axis was applied to obtain isotropic data (approximately mm). The diffusion tensor elements at each voxel were determined by least-square fitting and diagonalized to obtain three eigenvalues and three eigenvectors. An eigenvector (e1) associated with the largest eigenvalue ( 1) was assumed to represent the local fiber direction. Anisotropy maps were obtained using the orientation-independent FA. Diffusion tensorbased color maps were created from this FA value (image intensity) and the eigenvector of the three vector elements. Colors on the map indicated fibers running along a right-to-left direction, anterior-to-posterior direction, and superior-to-inferior direction as red, green, and blue, respectively. Fiber tracking was then performed by means of a combination of Volume-one and dt1 (12). Briefly, tracking was initiated from a manually selected seed area, from which lines were propagated in anterograde and retrograde directions according to the eigenvector at each pixel. Because we were interested in drawing only the motor tracts in this study, one seed area for the CST and the other for the somatosensory tracts were placed on the cerebral peduncle and the ML, respectively, on the basis of anatomic knowledge of the fiber projections (14, 18). Tracking was terminated when it reached a pixel with FA lower than certain thresholds (range, ). These thresholds were first varied and subsequently determined to extract only the targeted tracts such as the CST. Tract reconstruction required approximately 1 minute using a 3-GHz pentium 4 workstation with 2 Gb of random access memory. Magnetoencephalography SEFs were recorded by applying rectangular electric pulses with duration of 0.2 milliseconds to the left or right median nerve. The stimuli were delivered with a constant interstimulus interval of 211 milliseconds at an intensity sufficient to generate a moderate thumb twitch. MEG was recorded using a 204-channel biomagnetometer (VectorView; Elekta Neuromag Oy, Helsinki, Finland) in a magnetically shielded room, and the SEF recordings consisted of a 50-millisecond prestimulus baseline and a 250-millisecond analysis period after stimulus delivery. MEG responses of 200 epochs were averaged and digitally filtered between 1 and 70 Hz. Deflections of SEFs were visually identified on the basis of the root meansquared fields of more than 10 sensors in the hemisphere contralateral to the stimulus. Equivalent single current dipoles were calculated for the field data that peaked at approximately 20 and 30 milliseconds. We accepted only the dipoles with a correlation value of more than 0.95 and a confidence volume of less than 200 mm 3. The coordinates of the dipoles were transferred to the anatomic MRI scans by identifying external anatomic fiduciary markers (i.e., nasion, left or right preauricular points). The estimated SEF dipoles were projected onto the reconstructed brain surface images, clearly indicating the CS and PMA locations. Intraoperative Inspection of the Brain Surface and the Cortical SEP Recording One of the programs equipped with the MEG system created anatomic MRI slices, including the SEF dipoles, with a digital imaging and communications in medicine format. The created MRI data with dipoles were transferred via fast ethernet to a neuronavigation system (Stealthstation treatment guidance platform; Medtronic Sofamor Danek, Inc., Memphis, TN) (9) and used during surgery. To confirm the CS location predicted by functional imaging techniques, an intraoperative phase reversal of the cortical SEPs was measured with a single electrode strip containing four electrodes. For stimulation, we applied 200 repetitions of 0.2-millisecond constant current pulses delivered to the medial nerve at the wrist, with a frequency of 5.1 Hz and a current strength of 10 to 20 ma. The electrode was placed on several different places, including the suspected primary somatosensory area, PMA, and surrounding areas before resection of a lesion, comparing the actual brain surface in the operative field and the brain surface MRI scans. We investigated the relationship between the locations of the CS, a SEP electrode demonstrating the highest amplitudes of N20 (active electrode), and the SEF dipoles on the neuronavigation system. Furthermore, the brain surface MRI scans, including the results of the corticospinal tractography NEUROSURGERY VOLUME 56 OPERATIVE NEUROSURGERY 1 JANUARY 2005 ONS-101

5 KAMADA ET AL. FIGURE 1. Three-dimensional T1-weighted MRI scans (A), DTI scans (B), and combined images of T1-weighted MRI and DTI (C) showing no critical image distortions on DTI, except for areas surrounding the frontal sinus and air cells. T1-weighted MRI and DTI were a good fit in the upper brainstem and rolandic regions. and the SEF dipoles, were visually compared with intraoperative findings, observing the key sulci and the lesions. RESULTS fmri Six patients (Patients 3, 9, 14, 17, 18, and 27) could not perform the self-paced finger-tapping task because of hemiparesis, claustrophobia, or dementia. Motor evoked responses, mainly in the suspected PMA, were revealed in 22 cases. Multiple areas of the motor evoked responses were observed in the contralateral supplementary motor area (SMA), the cingulate gyrus, and the ipsilateral and contralateral somatosensory motor areas in two patients (Patients 23 and 26). Thus, the CS and PMA were identified by fmri in 22 (73.3%) of 30 cases. MEG Because of rapid neurological deterioration requiring surgical treatment, three patients (Patients 14, 15, and 17) were not available for the MEG investigations. Patient 27, who had a meningioma associated with severe perifocal edema in the parietal region and hemisensory deficit, demonstrated no typical SEF deflections. Twenty-six (96.2%) of 27 patients who underwent MEG investigation typically revealed two SEF deflections (N20m and P30m), and the estimated SEF dipoles were located in the inner bank of the CS. Corticospinal Tractography There were no critical image distortions on DTI scans, except in the areas surrounding the frontal sinus and air cells. The anatomic MRI and DTI scans were a good fit after the image registration in all patients (Fig. 1). Figure 2 shows two representative slices of diffusion tensorbased (grayscale and color) FA maps compared with histological preparations and anatomic illustrations of the tract in the brainstem. The diffusion tensor-based maps grossly delineate a complex substructure within the white matter of the brainstem. The brightness of the grayscale maps, which reflects the magnitude of anisotropy, provides high contrast between the white and gray matter, whereas the color, which indicates the orientation of tracts, differentiates various tracts. In particular, major tracts such as the CST (purple), ML (bluepurple), and middle cerebellar peduncle (green) are identified on the diffusion tensor-color maps. The CST runs in the cerebral peduncle of the midbrain and is separated from the ML by the substantia nigra, although both tracts longitudinally run along the craniocaudal direction. A diffusion tensor-color map of the midbrain delineates the CST and ML as a purple area in the cerebral peduncle and a blue-purple area in the tegmen, respectively. Conversely, the CST is diffusely spread in various directions, and there are no boundary structures between the CST and ML in the pons. The diffusion tensorcolor map of the pons thus reveals these tracts as mixed colors (red and purple) and hardly distinguishes the CST from other tracts. The middle cerebellar peduncle is demonstrated as a green area in the lateral part of the pons. The algorithm for identifying and mapping tracts was fully automated but required a seed area. After the image registration, the corticospinal tractography results were precisely fused with the anatomic MRI. Figure 3 demonstrates representative samples for the motor and somatosensory tracts in two normal subjects. It was possible to delineate the motor and somatosensory trajectories separately as red and blue lines, taking seed areas of the CST (red circle) and ML (blue circle) on the midbrain, respectively. The red seed area was always placed at the cerebral peduncle, containing only the CST as shown in Figure 3. The motor tracts (red lines) consistently propagated from the cerebral peduncle to the suspected PMA location. As long as the red seed area excluded the ML on the color map, the results of the tractography were reproducible. The blue seed area (semicircular shape) was placed in the tegmen of the midbrain for the ML (Fig. 3). For visualization of tracts of interest, it was most crucial to place a seed area of ONS-102 VOLUME 56 OPERATIVE NEUROSURGERY 1 JANUARY

6 CORTICAL MAPPING BY TRACTOGRAPHY FIGURE 3. MRI scans showing the results of tract tracking superimposed on the three-dimensional T1-weighted images using colored tracts such as the CSTs (red) and somatosensory tracts (blue) in two patients (A and B). They are selectively visualized by placing the seed areas on the cerebellar peduncle (red circle) and ML (blue semicircle), respectively. the PMA, the corticospinal tractography consistently demonstrated the dominant CST connection between the cerebral peduncle and PMA. In two patients (Patients 14 and 18) with glioma in the frontal lobe, the CST partly lost the specific anisotropy as a result of tumor invasion. It was thus necessary to take a low FA (0.25) to find the continuous CST trajectories to the PMA. As a result, we could indicate the PMA location by corticospinal tractography. FIGURE 2. Comparison of histological findings (A and E), drawings (B and F), and DTI-based gray (C and G) and color (D and H) maps showing the midbrain and the pons. In the drawings, the red dots indicate the descending fibers. The intensity of the DTI gray maps is scaled in proportion to the degree of diffusion anisotropy. On the DTI color maps, red, green, and blue indicate fibers running in the right-to-left, ventrodorsal, and craniocaudal directions, respectively. The CST (purple), ML (bluepurple), and middle cerebellar peduncle (MCP, green) are identified. interest on an appropriate region, avoiding contamination of adjacent tracts such as the middle cerebellar peduncle, ML, or other tracts. The FA threshold should be carefully determined for the tracking processes. In this study, when the FA threshold approached that of gray matter ( 0.25), tracking of the CST and ML started to include adjacent tracts. Between the FA thresholds of 0.25 and 0.35, the tracking reproduced the same trajectories with respect to length and direction. The sole difference was the diameter of the trajectories, which depended on FA. Regarding identification of Intraoperative Findings Cortical SEPs were measured in the affected hemisphere of all patients and clearly indicated the CS locations. After the CS identification by cortical SEPs, we could easily recognize the key sulci, comparing the surgical fields and the preoperative brain surface MRI. The neuronavigation system, including the SEF dipoles, was helpful to identify locations of the active SEP electrode and the CS. It was practical to observe the sulcal patterns of the surgical fields by visual inspection, however, compared with the preoperative brain surface MRI. The CS locations identified by the cortical SEPs and the corticospinal tractography were consistently identical in all of the patients. The results of functional mapping are summarized in Table 2. Illustrative Cases Patient 18 Patient 18 was a 49-year-old right-handed woman presenting with left hemiparesis 2 months before the investigations. Radiological examinations revealed a heterogeneous high-intensity mass in the right NEUROSURGERY VOLUME 56 OPERATIVE NEUROSURGERY 1 JANUARY 2005 ONS-103

7 KAMADA ET AL. TABLE 2. Summary of identification of the central motor region in the affected hemispheres by the functional mapping techniques a Patient no. MEG fmri DTI Cortical SEPs 1 L L L L 2 L L L L 3 L Failed L L 4 L L L L 5 L L L L 6 L L L L 7 L L L L 8 L L L L 9 L Failed L L 10 L L L L 11 L L L L 12 L L L L 13 L L L L 14 ( ) Failed L L 15 ( ) L L L 16 L L L L 17 ( ) Failed L L 18 L Failed L L 19 L L L L 20 L L L L 21 L L L L 22 L L L L 23 L Multiple L L 24 L L L L 25 L L L L 26 L Multiple L L 27 Failed Failed L L 28 L L L L 29 L L L L 30 L L L L a MEG, magnetoencephalography; fmri, functional magnetic resonance imaging; DTI, diffusion tensor imaging; SEPs, somatosensory evoked potentials; L, central sulcus successfully localized; NL, central sulces not successfully localized; Failed, investigations failed because of patients conditions, such as cortical dysfunction; ( ), not available. frontal lobe on T2-weighted MRI scans (Fig. 4A), with ring-like enhancement of gadolinium diethylenetriamine penta-acetic acid on T1-weighted MRI scans, strongly suggesting a high-grade glioma. The estimated SEF dipoles clearly indicated the CS location on an axial section and a three-dimensional brain surface image (Fig. 4, C and E). The location of the active SEP electrode was exactly same as that of the SEF dipole on the neuronavigation system and confirmed the CS location in the surgical field (Fig. 4F). The patient could not complete the finger-tapping task because of severe hemiparesis; therefore, fmri was not performed. The tumor itself had low anisotropy, and the FA map demonstrated that the low-anisotropic lesion deformed the white matter structures and partly involved the CST. Using a relatively lower FA value (0.25), we could reconstruct the CST trajectory in the affected hemisphere. The CST tracking process demonstrated that the midbrain CST propagated toward the PMA, indicating the PMA location. The CS and PMA locations as determined by SEF and tractography on the three-dimensional brain surface image and those of the surgical fields were exactly identical (Fig. 4, E and F). With a short acquisition time and no requirement of active tasks, the obtained corticospinal tractography could provide the anatomic information on the CST trajectory and confidently localize the PMA. Tractography with diffusion tensor calculation could theoretically eliminate the isotropic components of the pathological brain tissue and trace the anisotropic components. After subtotal removal of the tumor, the patient s hemiparesis was slightly improved, and the histological diagnosis was glioblastoma multiforme. Patient 26 Patient 26 was a 16-year-old girl who experienced repeated transient numbness and weakness in the right hand. Radiological examination revealed a large intracranial mass in the left rolandic area associated with multilobular cysts. The main part of the tumor was hyperintense on T2-weighted MRI scans and isointense with marked enhancement with gadolinium diethylenetriamine penta-acetic acid on T1-weighted MRI scans. Because the lesion was adjacent to the PMA and compressed the CST, it was important to identify the motor system precisely. Although the estimated SEF dipoles were shifted downward and posteriorly by the tumor, they clearly indicated the CS location (Fig. 5, A and B). The patient clumsily performed the right finger-tapping task in the fmri examination because of mild hemiparesis. Motor evoked responses widely appeared in the contralateral (left) somatosensorymotor area, the SMA, and the ipsilateral (right) frontal lobe. The fmri result hardly indicated the PMA and CS locations (Fig. 5C). Conversely, the three-dimensional relationship between the CST and the tumor was clearly visualized on tractography. Despite mechanical CST compression by the tumor, the CST precisely reached the suspected PMA (Fig. 5, A and B). The CS location was confirmed by the N20 phase reversals (Fig. 5E). The location of the active electrode was the same as that of the SEF dipole on the neuronavigation system. The anatomic and functional orientations on the preoperative brain surface MRI exactly fit with the intraoperative inspection (Fig. 5, D and E). After surgery, the patient s hemiparesis immediately disappeared, and the histological diagnosis was meningioma. Patients 24 and 30 The locations of the CST relating to a mass lesion were various. Figure 6 shows a right parietal cavernous hemangioma (Fig. 6A, Patient 24) and a glioblastoma in the left parietal lobe (Fig. 6B, Patient 30). The CST is located anterior to the lesions. The anatomic relations between a lesion and the CST in the deep white matter were clearly demonstrated in all cases investigated. ONS-104 VOLUME 56 OPERATIVE NEUROSURGERY 1 JANUARY

8 CORTICAL MAPPING BY TRACTOGRAPHY FIGURE 4. A, T2-weighted MRI scan of Patient 18 showing a diffuse and heterogeneous high-intensity mass in the right hemisphere. B, FA map demonstrating that anisotropic components were partially involved but still preserved at the posterior border of the tumor. C and D, corticospinal tractography (right, orange; left, blue) and dipoles (white squares) of the SEFs showing the location of the PMA and CS (white arrow). E, three-dimensional brain surface MRI scan showing the location of the PMA and indicating the corticospinal tractography and CS location (arrows) by means of a SEF dipole. F, intraoperative photograph of the brain surface showing the CS, confirmed by cortical SEPs (arrows). The location of the electrode showing the highest N20 amplitude was exactly same as that of the SEF dipole on the neuronavigation system. The simulated operative field shows the geographic relationship between corticospinal tractography, SEF dipoles, and the expected CS location. DISCUSSION In this study, the corticospinal tractography with 0.35 of FA and the appropriate selection of a seed area in the cerebral peduncle clearly indicated the PMA location in 28 of 30 patients; in the remaining two cases, with marked tumor invasion to the precentral gyrus, a reduced level (0.25) of FA was able to indicate the PMA. Corticospinal tractography is thus considered to be an effective technique to identify the PMA location noninvasively, although it reflects only the anisotropic components of the axonal fascicles in the subcortical structures. In addition, from a clinical point of view, it is important to complete the data acquisition process rapidly and easily without requiring any tasks of a patient. It was achieved within a few minutes, applying the method described in this report. DTI is a technique that can characterize the spatial properties of molecular diffusion processes. Because a fascicle is composed of collections of similarly oriented axons, it generally exhibits high anisotropy values (3, 16). Diffusion within each voxel may be most naturally described by three mutually perpendicular eigenvectors, whose magnitude is indicated by three corresponding eigenvalues. The eigenvalues are the three principal diffusion coefficients measured along three (intrinsic) eigenvector directions, which define the local fibers frame of reference for each voxel. The direction of the fibers is thus indicated by the eigenvector of the largest eigenvalue of the diffusion tensor (3). Tractography searches the direction and strength of anisotropy in each pixel and visualizes the tract profiles, including the subcortical fibers. A number of previous reports have described the relationship between brain lesions and subcortical fiber connections (8, 15 17, 21). FA is one of the indices for anisotropy and might reflect the fascicles integrity. Stieltjes et al. (20) have demonstrated the appropriate FA values (range, ) for drawing tractography of major fascicles in the brain. When varying the FA, the number of voxels involved decreased with increasing FA threshold. We adopted a minimum FA value (0.25) to trace the CST in severely affected brain. Tractography, which is the result of a technique using complex mathematical principles, may involve untargeted fiber bundles. The spinothalamic tracts should terminate in the thalamus. In our study, however, the somatosensory tracts defined in the ML of the midbrain propagated toward the anterior part of the parietal lobe via the thalamus. The further propagation of the somatosensory tracts beyond the thalamus thus may involve anisotropic components of other fibers following the spinothalamic tract. For clinical application of this technique, it is necessary to exploit the obtained result of tractography carefully and appropriately. NEUROSURGERY VOLUME 56 OPERATIVE NEUROSURGERY 1 JANUARY 2005 ONS-105

9 KAMADA ET AL. FIGURE 5. Sagittal and transverse T1-weighted MRI scans of Patient 26 superimposed on the SEF dipoles (white square) and corticospinal tractography (A and B) showing the location of the CS (white arrow) and PMA. C, fmri scan demonstrating multiple activations in the contralateral and ipsilateral frontal regions as well in as the SMA during the right finger-tapping test. D, three-dimensional brain surface MRI scan, including the SEF dipoles, showing the suspected locations of the CS (arrows). E, photograph of the surgical field, including the electrode locations of cortical SEPs, showing the CS. The intraoperative inspection and monitoring confirmed the CS location by observation of the N20 phase reversals between Electrodes 2 and 3. The location of the electrode showing the highest N20 amplitude was exactly same as that of the SEF dipole on the neuronavigation system. The main purpose of this study was to identify the PMA using corticospinal tractography. We paid special attention to placing a seed area and think that the midbrain is an appropriate level for the CST seed areas, because the cerebral peduncle of the midbrain has a unique anatomic structure in which the motor (CST) and somatosensory tracts are completely separated and those fibers are densely packed. Furthermore, there was less susceptibility artifact of DTI at the midbrain than at the pons or the medulla oblongata. When the seed areas were placed in the internal capsule or the lower brainstem, where the motor tract is anatomically close to the somatosensory tract, there was no guarantee that the reconstructed tracts reflected only the CST. In contrast to the CST in the cerebral peduncle, it was hard to enclose the ML exclusively as a seed area because of its anatomic size and localization. This might be another reason for the extended propagation of the somatosensory tracts to the parietal white matter. Although the CST is the largest axonal bundle in the human brain, the tracts with a craniocaudal direction in the cerebral peduncle originate from various cortices such as the frontal and parietal lobules (14, 18). Maier et al. (13) made a quantitative comparison of the density of the CST projections from the PMA and SMA to the spinal motor nuclei in macaque monkeys. Densitometric analysis revealed that the CST projections from the PMA were much denser than from the SMA. In the caudal Th1, the densest projections from the PMA occupied 81% of this motoneuronal area compared with only 6% from the SMA. Furthermore, Jane et al. (7) studied the contribution of the PMA to the CST in a patient with involuntary movement and concluded that the CST is composed of more than 10 6 fibers, originating almost exclusively from the PMA. In addition, it has been reported that only approximately 20% of the CST originates from the somatosensory cortex (4). These facts prove that the connections between the cerebral peduncle and PMA are extremely predominant in the CST at the midbrain. The appropriate seed areas and the FA values enable us to extract the targeted tracts. The results of corticospinal tractography presented here thus seem to indicate the CS and PMA locations. The widespread use of MRI scanners made blood oxygen level-dependent fmri an ideal tool for functional mapping, but the true origin of the gradient echo fmri signal is under discussion. The activation signals of fmri studies are 1 to 8% greater than the noise using clinical 1.5-T MRI scanners. The source of the blood oxygen level-dependent fmri signals may have a shift from the activated brain area, depending on pathological brain conditions and venous drainage architecture (1, 6, 8). If the autoregulation of the cerebral vessels is lost in brain tissue that still functions, the affected region may hardly respond to increased neural activity by a corresponding increase in blood flow (1, 6). Therefore, certain pathophysiological conditions and venous drainage architecture may strongly interfere with the cortical mapping of fmri. ONS-106 VOLUME 56 OPERATIVE NEUROSURGERY 1 JANUARY

10 CORTICAL MAPPING BY TRACTOGRAPHY major tracts in the white matter. It is still unable to differentiate between tract projections for hand, foot, and other body areas. The spatial resolution of this diffusion tensor study was, however, sufficient to localize the PMA through visualization of the CST. In the near future, several new techniques should improve the quality of tractography, such as high angular sampling of the k space, interpolation or regularization of the tensor field, and global energy minimization for tract tracking. CONCLUSION Tractography is a rapid and noninvasive method of brain mapping using routine MRI scanners. We think that this technique is likely to become one of the major methods for identifying the PMA, especially in patients who have severe cortical dysfunctions such as paresis, sensory deficit, and consciousness disturbance. REFERENCES FIGURE 6. Transverse and sagittal T1-weighted MRI scans of Patients 24 (A) and 30 (B) with lesions in the parietal lobe. The corticospinal tractography indicates the location of the CS (white arrows) and PMA. A MEG localization error originates from the MEG data acquisition of evoked magnetic fields and dipole localization when using the single equivalent current dipole model and a spherical head model as a volume conductor. Another source of error is introduced when the dipole localizations in the MEG coordinate system are transferred to the MRI data set, because fiduciary marker placement and the registration process can affect accuracy. Although the overall accuracy of the combination of MEG localization is approximately 5 mm, it is clinically acceptable (5, 14). MEG is available at only a limited number of institutions because of its cost, whereas tractography has a greater possibility of becoming the prevalent method. Echo-planar imaging is a recently established technique. Its ultrafast scanning can minimize motion artifact and is indispensable for fmri and DTI. A critical issue with this technique is the geometric distortion induced by magnetic field heterogeneity and motion probing gradient. To minimize scanning artifact, we kept the magnetic field of our MRI systems as homogeneous as possible and applied morphing compensation after data acquisition. As a result, the reconstructed DTI scans were a good fit to the anatomic MRI scans (Fig. 1). The fiber orientation reflects the average orientation of axonal fibers in each pixel and is susceptible to a grade of tissue heterogeneity. Within a pixel, numerous fibers are always crossing and a few of the fibers have a different orientation from the dominant fibers. Therefore, at the present time, tractography can only provide gross anatomic information about 1. Alkadhi H, Kollias SS, Crelier GR, Golay X, Hepp-Reymond MC, Valavanis A: Plasticity of the human motor cortex in patients with arteriovenous malformations: A functional MR imaging study. AJNR Am J Neuroradiol 21: , Ardekani BA, Braun M, Hutton BF, Kanno I, Iida H: A fully automatic multimodality image registration algorithm. J Comput Assist Tomogr 16: , Basser PJ, Mattiello J, LeBihan D: MR diffusion tensor spectroscopy and imaging. Biophys J 66: , Coenen VA, Krings T, Axer H, Weidemann J, Kraenzlein H, Hans FJ, Thron A, Gilsbach JM, Rohde V: Intraoperative three-dimensional visualization of the pyramidal tract in a neuronavigation system (PTV) reliably predicts true position of principal motor pathways. Surg Neurol 60: , Ganslandt O, Fahlbusch R, Nimsky C, Kober H, Moller M, Steinmeier R, Romstock J, Vieth J: Functional neuronavigation with magnetoencephalography: Outcome in 50 patients with lesions around the motor cortex. J Neurosurg 91:73 79, Holodny AI, Schulder M, Liu WC, Wolko J, Maldjian JA, Kalnin AJ: The effect of brain tumors on BOLD functional MR imaging activation in the adjacent motor cortex: Implications for image-guided neurosurgery. AJNR Am J Neuroradiol 21: , Jane JA, Yashon D, DeMyer W, Bucy PC: The contribution of the precentral gyrus to the corticospinal tract of man. J Neurosurg 26: , Kamada K, Houkin K, Iwasaki Y, Takeuchi F, Kuriki S, Mitsumori K, Sawamura Y: Rapid identification of the primary motor area by using magnetic resonance axonography. J Neurosurg 97: , Kamada K, Houkin K, Takeuchi F, Ishii N, Ikeda J, Sawamura Y, Kuriki S, Kawaguchi H, Iwasaki Y: Visualization of the eloquent motor system by integration of MEG, functional, and anisotropic diffusion-weighted MRI in functional neuronavigation. Surg Neurol 59: , Kamada K, Oshiro O, Takeuchi F, Kuriki S, Houkin K, Iwasaki Y, Abe H: Identification of central sulcus by using somatosensory evoked magnetic fields and brain surface MR images: Three dimensional projection analysis. J Neurol Sci 116:29 33, Kamada K, Takeuchi F, Kuriki S, Oshiro O, Houkin K, Abe H: Functional neurosurgical simulation with brain surface magnetic resonance images and magnetoencephalography. Neurosurgery 33: , Kunimatsu A, Aoki S, Masutani Y, Abe O, Mori H, Ohtomo K: Threedimensional white matter tractography by diffusion tensor imaging in ischaemic stroke involving the corticospinal tract. Neuroradiology 45: , Maier MA, Armand J, Kirkwood PA, Yang HW, Davis JN, Lemon RN: Differences in the corticospinal projection from primary motor cortex and supplementary motor area to macaque upper limb motoneurons: An anatomical and electrophysiological study. Cereb Cortex 12: , NEUROSURGERY VOLUME 56 OPERATIVE NEUROSURGERY 1 JANUARY 2005 ONS-107

11 KAMADA ET AL. 14. Mesulam MM: Tracing Neural Connections with Horseradish Peroxidase. Chichester, Wiley, Mori S, van Zijl PC: Fiber tracking: Principles and strategies A technical review. NMR Biomed 15: , Mori S, Frederiksen K, van Zijl PC, Stieltjes B, Kraut MA, Solaiyappan M, Pomper MG: Brain white matter anatomy of tumor patients evaluated with diffusion tensor imaging. Ann Neurol 51: , Nimsky C, Ganslandt O, Kober H, Moller M, Ulmer S, Tomandl B, Fahlbusch R: Integration of functional magnetic resonance imaging supported by magnetoencephalography in functional neuronavigation. Neurosurgery 44: , Orioli P, Strick P: Cerebellar connections with the motor cortex and the arcuate premotor area: An analysis employing retrograde transneuronal transport of WGA-HRP. J Comp Neurol 22: , Roberts TP, Disbrow EA, Roberts HC, Rowley HA: Quantification and reproducibility of tracking cortical extent of activation by use of functional MR imaging and magnetoencephalography. AJNR Am J Neuroradiol 21: , Stieltjes B, Kaufmann WE, van Zijl PC, Fredericksen K, Pearlson GD, Solaiyappan M, Mori S: Diffusion tensor imaging and axonal tracking in the human brainstem. Neuroimage 14: , Wieshmann UC, Symms MR, Parker GJ, Clark CA, Lemieux L, Barker GJ, Shorvon SD: Diffusion tensor imaging demonstrates deviation of fibers in normal appearing white matter adjacent to a brain tumor. J Neurol Neurosurg Psychiatry 68: , Acknowledgments We thank Drs. Osamu Takizawa and Jun Okamoto at Siemens-Asahi (Shinagawa, Japan) for technical support. This work was supported in part by a Research Grant for Cardiovascular Disease from the Ministry of Health and Welfare of Japan. COMMENTS The authors of this article used corticospinal tractography to identify the location of the central sulcus and primary motor area (PMA) in 30 patients with mass lesions affecting the primary motor cortex. For validation of their method, they identified the location of the central sulcus and PMA, primarily using intraoperative phase reversal of the cortical somatosensory evoked potential (SEP) and, secondarily, by functional magnetic resonance imaging (fmri) and magnetoencephalography (MEG). The central sulcus locations, identified by corticospinal tractography and the cortical SEP, were identical in all 30 patients. Furthermore, the authors were able to identify the location of the central sulcus and PMA in 22 of 30 patients with fmri and 26 patients with MEG from a group of 27 who were able to undergo MEG. It has been known for some time that the diffusion of free water molecules is not the same in all directions of three-dimensional space in white matter; instead, it is anisotropic. The diffusivity of water is found to be greatest along the dominant orientation of white matter tracts and is influenced by its micro- and macrostructural properties such as intra-axonal organization, degree of myelination, individual fiber diameter, fiber density, and neuroglial cell packing. This knowledge has led to a variety of methods for displaying fiber orientation, ranging from simple techniques based on apparent diffusion coefficients to methods based on information contained within the full diffusion tensor. The diffusion tensor defines a principal frame of directions for each voxel by its eigenvectors. A tensor is a mathematic construct that describes the properties of an ellipsoid in three-dimensional space. MRI diffusion tensor imaging (DTI), developed by Basser et al. (1), has provided a means of harnessing the Brownian motion property or diffusivity of water molecules in vivo for tractography. Tractography reconstructs white matter tracts in three-dimensional space using the anisotropic diffusion of free water molecules within white matter tracts. This technique is dependent on precise estimation of the orientation of the tensor within each voxel. There are a few potential limitations to this method that should be kept in mind. First, the direction of the measured principal eigenvector is determined on the basis of a voxel average and does not necessarily represent the trajectories of individual microscopic tracts. The voxel sizes used in DTI are significantly larger than the white matter structures that are being mapped. This leads to significant partial volume errors and may lead to failure to identify fiber trajectory when a voxel is made up of nonuniformly distributed fibers. Second, image noise influences the direction of the major eigenvector, leading to directional error and its accumulation in the estimated fiber tracts. Third, as the degree of anisotropy decreases, the error in calculation of the major eigenvector increases, which could result in erroneous tracking in regions in which the diffusion tensor does not have strong directional property, such as the thalamus. In their study, the authors contend that the somatosensory tracts defined in the medial lemniscus of the midbrain propagated beyond the thalamus and toward the anterior part of the parietal lobe, which could be the result of this type of error. Fourth, the choice of position, shape, and density of the seed point as well as the anisotropy threshold is a key step in precise mapping of the tracts and is currently user defined and not yet automated or standardized. White matter tractography based on DTI is a rapidly evolving method of central nervous system imaging, with many challenges and exciting new applications in fmri such as identifying the nuclei of the thalamus. Practical routine application of this technology requires fiber-tracking algorithms that can resolve fiber heterogeneity in a voxel to define crossing and dispersing white matter tracts accurately. Furthermore, imaging using highresolution voxels should reduce the error introduced by volume averaging in the direction of the principal eigenvector. This would lead to a better representation of the actual orientation of the fiber tract within a voxel as well as to better accuracy of the tracking algorithm. The authors of this article have done a reasonable job in using this technology to tackle an important problem, namely, preoperative identification of the central sulcus and PMA in patients with mass lesions affecting these regions, and have successfully validated their method using the intraoperative cortical SEP. Farhad M. Limonadi Kim J. Burchiel Portland, Oregon 1. Basser PJ, Pajevic S, Pierpaoli C, Duda J, Aldroubi A: In vivo fiber tractography using DT-MRI data. Magn Reson Med 44: , One of the most important tasks for the neurosurgical oncologist operating near the motor areas is correct identification of motor fibers. There are six techniques now used to aid in the ONS-108 VOLUME 56 OPERATIVE NEUROSURGERY 1 JANUARY