Intraoperative localization of subcortical brain lesions

Acta Neurochir (Wien) (2008) 150: 537–543 DOI 10.1007/s00701-008-1592-z Printed in The Netherlands

Clinical Article Intraoperative localization of subcortical brain lesions V. Esposito1 , S. Paolini2 , R. Morace1 , C. Colonnese3 , E. Venditti3 , V. Calistri3 , G. Cantore1 1

Universita degli Studi di Roma ‘‘Sapienza’’, Cattedra di Neurochirurgia – IRCCS Neuromed, Pozzilli, Italy Universita degli Studi di Perugia, Cattedra di Neurochirurgia – IRCCS Neuromed, Pozzilli, Italy 3 Universita degli Studi di Roma ‘‘Sapienza’’, Cattedra di Neuroradiologia – IRCCS Neuromed, Pozzilli, Italy 2

Received 27 August 2007; Accepted 26 March 2008; Published online 6 May 2008 # Springer-Verlag 2008

Summary Background. Some brain tumors may grow immediately beneath the cortical surface without distorting its appearance. Intraoperative image guidance promotes safe resection. We have developed MRI-based corticotopography (MRI-bct), to localize lesions during surgery, using simple, non-dedicated equipment, to match a three-dimensional reconstruction with the corresponding appearance of the brain cortex. Methods. Forty-six patients underwent resection of subcortical brain lesions, aided by MRI-bct. The lesions had a maximum diameter less than 3 cm, were subcortical but no deeper than the floor of the nearest cerebral sulcus. Each patient had a volumetric MRI scan with and without contrast administration. Data sets were transferred to a laptop personal computer and processed using a rendering software. At operation, the three-dimensional model of the brain, including a surface overlay of the lesion, was matched to the exposed brain surface. After its exact relationship with the overlying sulcal pattern was defined, the lesion was localized and resected. In selected patients, the procedure was coupled with functional brain mapping. Results. Data processing took from 10 to 15 min and could be done whenever convenient before operation. Surface matching between the surgical field and the reformatted MRI always required less than 5 min and

Correspondence: Dr. Sergio Paolini, Via G.A. Sartorio 44, 00147 Rome, Italy. e-mail: [email protected]

was done near the operating table. In all patients, the lesion was identified at the first attempt, through a small corticotomy, regardless of the brain shift after dural opening. Conclusions. MRI-bct is a practical, time-saving neuronavigational aid ideal for localizing superficial lesions underlying the cerebral cortex because it unmistakably characterizes the adjacent sulcal anatomy. Keywords: Brain cortex; MRI; neuronavigation; reformatting; three-dimensional.

Introduction Subcortical brain tumors may reach a considerable size, yet be hidden from view on the surface of the brain. These lesions are usually exposed using intraoperative image-guidance [5, 6, 15] with magnetic resonance imaging (MRI)-based workstations. These systems nevertheless have minor drawbacks including lengthy operating times, costs, and loss of accuracy owing to ‘‘brain shift’’ [11]. Although real-time navigation systems such as intraoperative MRI or three-dimensional (3D) ultrasonography are viable alternatives [8, 10, 14], they also require dedicated equipment. In 2002 we developed an inexpensive technique for combining information from preoperative MRI and intraoperative appearances, MRI-based corticotopography (MRI-bct). The technique incorporates a few steps aimed at simplifying the surgeon’s orientation around the cerebral cortex and was initially used for topographical

538

V. Esposito et al.

localization of subdural grid electrodes implanted in epileptic patients. Subsequently, encouraged by the results and by how straightforward the procedure proved, we extended its application to operation on brain lesions [4]. MRI-bct is now part of our routine approach to many intracranial diseases. We describe its use in a series of 46 patients undergoing surgical resection of a subcortical brain lesion.

Technique Before operation, a volumetric MRI image is acquired before and after intravenous injection of gadolinium (0.4 cc=kg), with a 1.5 Tesla Signa scanner (General Electric, Milwaukee, WI – T1 axial sequences, with the following parameters: repetition time ¼ 8.9 msec; echo time ¼ 1.8 msec; flip angle ¼ 20
; 232 axial slices, slice thickness ¼ 1.4 mm; number of averages 2; field of view (FOV) ¼ 260.00 mm; and matrix size 256  256). The MRI data sets, acquired in digital imaging and communication (DICOM) format, are transferred to a laptop personal computer using a USB memory unit and processed using MRIcro, a software program freely available on the Internet [12]. The MRI data sets are first converted into a single multislice ‘‘Analyze’’ file. The native DICOM images, originally acquired as anisotropic scans, are then reformatted into isotropic voxels. Next, to extrapolate the brain parenchyma, the ‘‘brain extraction tool’’ (BET) interactively segments the MRI image outline, isolating the brain from the overlying structures. The extracted data sets are then processed for 3D rendering. The resulting 3D model of the brain can be grasped and rotated according to the required perspective. Depending on the surgeon’s needs, a subcortical lesion may be outlined, reconstructed and projected as ‘‘region of interest’’ (ROI) over superficial structures. Before operation, the bone flap can be planned by projecting the ROI over the external bone and over the skin and using cranial sutures, temporal lines, craniometric points, the mastoid bone and the external ear as surface landmarks. If required, greater precision can be obtained by applying one or two fiducials over the skin approximately corresponding to the ROI projection area and including these artificial landmarks into the threedimensional MRI. After dural opening, the ‘‘real’’ cortical surface and its virtual reconstruction, including a cortical ROI projection, must be compared. Usually, the surgeon can immediately compare them directly. However, once the brain is exposed, we routinely obtain a high-resolution

Fig. 1. Intraoperative picture showing the surface matching process, done in the operating theater with a personal computer

digital photograph of the cortex and immediately transfer it to the computer where the reformatted images are stored. One of the commonly available photo-editing software programs is then used to crop, rotate and compare the intraoperative photo with the 3D reconstruction of the cortex. This step makes it easy to see how the sulcal patterns in the photo and the reconstruction correspond (Fig. 1). The cortical ROI projection can then be traced on the operative field, thus allowing selective corticotomy. If the MRI reconstructions show that the lesion involves eloquent cortical areas, functional brain mapping is added. Application and results Between September 2004 and June 2006, 46 patients underwent surgery for an intrinsic brain lesion which fulfilled two MRI criteria: a maximum diameter not exceeding 3 cm and the most superficial side lying underneath normal cortex but no deeper than the bottom of the most adjacent cerebral sulcus. There were 17 low-grade neuroepithelial tumors (WHO classification), 12 high-grade astrocytomas, 7 metastatic tumors and 10 cavernous hemangiomas, diagnosed by postoperative histological examination. The patients’ age ranged from 12 to 76 years. In 27 patients, the lesion was within or strictly associated with eloquent cortex. Nineteen lesions involved the Rolandic or perirolandic region and 8 involved language areas (Broca’s region in 3, Wernicke’s region in 5). The clinical presentation was with seizures in 24 patients, focal neurological symptoms in 10 patients and aspecific symptoms (headache, dizziness, etc.) in 12 patients. After MRI acquisition, a member of the surgical team processed the DICOM data sets on a laptop personal

Intraoperative localization of subcortical brain lesions

computer. The whole procedure took from 10 to 15 min for each patient and could be done at any convenient time before surgery. At surgery, the same personal computer used for processing MRI was used in the operating theatre for surface matching between the rendered model and the intraoperative photograph of the brain (Fig. 1). The average size of the bone flap was about 25 cm2. The matching process was always successful for cortical areas measuring 18–20 cm2 (about 4  5 cm, as a rule) or more, though even smaller exposures were sufficient for patients with unusual cortical details such as tortuous sulci or cortical vessels. In other patients, the dural opening had to enlarged to reach the required size. In all patients,

539

the procedure lasted less than 5 min, and unmistakably identified the anatomical structures overlying the target. 3D reconstructions of non-enhanced scans effectively visualized the sulcal pattern of the cortex, while contrast-enhanced scans were used to define the relationship between the ROI and the adjacent cortical vessels. In 7 patients, artefacts related to perilesional edema blurred the 3D representation of the gyral anatomy. In these circumstances, surface matching relied also on the cortical vascular anatomy. For 27 tumors lying close to or within eloquent cerebral areas, MRI-bct was coupled with functional brain mapping. The primary motor area was identified in the same operative setting by directly stimulating the cortex (bipolar stimulation; 5-pulse

Fig. 2. Case 1 a) Axial T2-weighted magnetic resonance imaging (MRI) scan showing a cavernous hemangioma (histological diagnosis) of the right parietal lobe. b) Three-dimensionally reformatted MRI, including a surface projection of the lesion (dark area). c) Intraoperative surface matching: a digital photograph of the cortex overlying the lesion (top) is compared with a cropped and properly oriented reconstruction of the same area (bottom) including a projection of the lesion (dark area). Note how well the two sulcal patterns correspond. d) The surgical field after removal of the cavernoma

540

trains; frequency: 60–200 Hz; amplitude: 2–20 mA; single pulse duration: 0.5–1 msec) and recording the peripheral motor evoked potential (MEP). Language areas were localized and investigated preoperatively, after implantation of subdural grid electrodes [9], by administering stimulation series of 3–10 sec (bipolar stimulation; 5-pulse trains; frequency: 50 Hz; amplitude: 0.4–15 mA; single pulse duration: 0.3 msec) during neurological and neuropsychological evaluation. Finally, the cortical entry point was selected after evaluating the superficial projection of the ROI and the stimulation thresholds for the adjacent areas.

V. Esposito et al.

All the lesions were successfully exposed at the first attempt through a cortical incision measuring from 13 to 25 mm. Postoperative MRI showed that 42 patients had undergone a radiologically total removal. Four patients with high-grade gliomas had a subtotal removal. The follow-up ranged from 1 to 24 months (mean 10.3 months). After operation, 11 (24%) of the 46 patients, each of whom had a tumor in an eloquent cortical areas, experienced new transient deficits or worsening of preexisting deficits. In each of these, follow-up examination after 30-days showed that symptoms had resolved completely. Minimal postoperative dysphasia was a

Fig. 3. Case 2 a) Sagittal post-contrast magnetic resonance imaging (MRI) scan showing a cystic lesion involving the parasagittal rolandica area, on the right side. b) Three-dimensionally reformatted MRI. The lesion itself is projected close to the midline. c) Intraoperative surface matching. Note that although hidden by a layer of normal cortex, the lesion can be localized by comparing the sulcal pattern of the exposed cortex (top) with the 3D MRI (bottom). The scans obtained after gadolinium injection clearly identify the two cortical veins crossing the surgical field over the region of interest (arrows). d) Top Direct cortical stimulation under electromyographic recording confirms the location of the motor strip over the lesion (points ‘‘1’’ to ‘‘6’’ elicit motor responses on biceps brachii, deltoid, quadriceps femoris and tibialis anterior). Point 7, overlying the anterior portion of the cyst, is functionally silent and will be used as the cortical entry point. Bottom The surgical field at the end of the procedure

Intraoperative localization of subcortical brain lesions

common finding (5 patients) for lesions resected within language areas and was considered to be due to postsurgical edema, since it typically reached its apex within 24–48 h and resolved after 5–7 days. One patient with a high-grade astrocytoma experienced permanent worsening of a preoperative hemiparesis. Illustrative patients Case 1 a 24-year-old woman with a 5-year history of partial seizures was admitted to our Institute with an MRI diagnosis of cavernous malformation involving the right parietal lobe. The lesion measured approximately 12 mm and was located subcortically, where it came into contact with the floor of the postcentral sulcus (Fig. 2a). Resection of the cavernoma was planned with the aid of volumetric MRI reconstructions (Fig. 2b). At operation, surface matching of the intraoperative photograph and 3D MRI including a ROI projection (Fig. 2c), allowed us to identify precisely the sulcus overlying the lesion. Through a small cortical incision, a typical cavernous angioma was rapidly exposed and resected along with its peripheral hemosiderin ring (Fig. 2d). The patient had an uneventful postoperative course and was discharged 4 days after surgery. Case 2 a 57-year-old woman underwent radiological investigation of recurrent headache. MRI disclosed a cystic lesion in the right Rolandic area. A mural nodule, slightly enhancing after contrast injection, was visible on the anterior wall of the cyst (Fig. 3a). The preoperative MRI images were reformatted using MRIcro software. On 3D images, the lesion outline projected superficially along the motor strip (Fig. 3b). At operation, the same area was identified through photographic surface matching (Fig. 3c). The cortical entry point was selected on the basis of functional brain mapping (Fig. 3d). The lesion was quickly exposed through a small incision on the anterior aspect, non-functional at electrical stimulation, of the ascending frontal gyrus. The remaining portion of the gyrus was functionally eloquent and was identified as primary motor area. A partially calcified mural nodule was completely resected from the anterior wall of the cyst (Fig. 3d). Postoperative MRI documented lesion removal . After operation, the patient was neurologically intact. Histological examination showed a gangliocytoma. Discussion The MRI-bct procedure we have developed has proved a straightforward technique that simplifies the surgeon’s

541

orientation around the cerebral cortex. MRI-bct has the distinct advantage of combining the two traditional steps used in conventional navigation, data processing and registration, into a simple, single procedure, thus shortening operating time. The MRI dataset is processed with a personal computer. Registration simply entails a visual comparison between the virtual and real representations of the cortical surface. Sulci, gyri and cortical vessels act at the same time as reference points and targets of the surgical approach. Given their surface location, these structures remain easily recognizable despite brain shift due to brain retraction or depletion of cerebrospinal fluid. The procedure is rapid and inexpensive. It lengthens operative times by no more than 5 min, considerably less than the time required for conventional navigation which in our experience, using the Medtronic Stealthstation Treo system, averages 15–20 min. The only hardware requirements are a laptop personal computer and a digital camera. The processing software, MRIcro, is currently freeware. In this study, extending our previous unpublished experience using MRI-bct to localize subdural electrodes in epileptic patients and as a navigational aid in intracranial diseases requiring surgical exploration of the brain surface [4] we focussed on patients with subcortical lesions lying no deeper than the bottom of the adjacent sulcus. We used these selection criteria because the sulcus itself serves as an intraoperative landmark to avoid an incorrect surgical trajectory. Patients with a deep-seated lesion or lesion located beneath unexposed cortical areas, usually the inferior and mesial surfaces of the cerebral hemispheres, are treated with standard neuronavigation. For deep-seated lesions, longer distances between the cortical surface and the surgical target would incur an inherent risk of error because, even after successful surface matching, the surgical trajectory remains uncontrolled. In the patients we selected according to our criteria, we successfully localized all the lesions at the first attempt. The postoperative morbidity we observed in our series was entirely limited to patients with a lesion in eloquent cortical areas. Overall, the occurrence of transient morbidity in a third of these compares well with larger series [3]. It seems that a small amount of postoperative edema can interfere with normal neurological functioning for some days. Others have already described the advantages of reformatted MRI in neurosurgery. Thus, Bastos and colleagues used three-dimensional MRI reconstructions for diagnostic investigation of medically intractable partial

542

epilepsy [1, 2]. Unlike us, they intended the procedure not as a navigation method but as a diagnostic tool, which they considered superior to conventional MRI in identifying subtle cortical abnormalities. Curvilinear reformatting of T1-weighted MRI data sets has been reported by Schulze-Bonhage et al. to simplify localization of subdural strip and grid electrodes implanted for preoperative evaluation of epileptic patients [13]. Although the information gained from the rendered images was used to tailor the operative resection, the procedure was not clearly described. More recently, Hattingen et al. used brain surface reformatted images to localize perirolandic lesions [7]. Their technique relied on post-processing of MRI data through a dedicated workstation, to produce a two-dimensional, unfolded view of the brain convexity. In their patients, functional MRI served also to confirm the position of the motor strip. The same authors underlined the main limitations of the technique: stretching the 3D anatomy into a planar view keeps the rolandic fissure recognizable but distorts the anatomical distances making them unlikely to fit the cortical anatomy exposed at operation. Furthermore, the technique can be applied only to the perirolandic region. The technique we describe overcomes these limitations because it aids both preoperative planning and operation itself. Moreover, using intraoperative digital photographs for surface matching allows the whole brain convexity to be explored in detail. Using this, even tiny sulci could be traced on the rendered MRI and spatially referred to the surgical target, thus optimizing the surgical trajectory. Factors limiting the range of application of this technique are inability to expose enough cortical surface, excessive depth of the target or significant brain edema, blurring the MRI representation of the gyral anatomy. The latter, as mentioned in the results section, can be overcome by relying on the vascular anatomy (contrastenhanced MRI) for surface matching. Future developments in this technique might include automatization of the matching process, and possibly integration within currently available surgical microscopes. Finally, because of its intrinsic accuracy in localizing superficial landmarks, MRIbct might be used to validate standard navigation results, quantifying the degree of brain shift error while exploring the brain surface.

Conclusions MRI-bct is a useful neuronavigational aid during operations on subcortical lesions. It enables the surgeon to

V. Esposito et al.

plan operation precisely using only a personal computer. Operating times seem considerably shorter than those required by conventional neuronavigation. The technique is applicable only to superficial lesions, those located deep or beneath unexposed cortical areas require conventional neuronavigational assistance.

References 1. Bastos AC, Comeau RM, Andermann F, Melanson D, Cendes F, Dubeau F, Fontaine S, Tampieri D, Olivier A (1999) Diagnosis of subtle focal dysplastic lesions: curvilinear reformatting from threedimensional magnetic resonance imaging. Ann Neurol 46: 88–94 2. Bastos AC, Korah IP, Cendes F, Melanson D, Tampieri D, Peters T, Dubeau F, Andermann F (1995) Curvilinear reconstruction of 3D magnetic resonance imaging in patients with partial epilepsy: a pilot study. Magn Reson Imaging 13: 1107–1112 3. Duffau H, Capelle L, Denvil D, Sichez N, Gatignol P, Taillandier L, Lopes M, Mitchell MC, Roche S, Muller JC, Bitar A, Sichez JP, van Effenterre R (2003) Usefulness of intraoperative electrical subcortical mapping during surgery for low-grade gliomas located within eloquent brain regions: functional results in a consecutive series of 103 patients. J Neurosurg 98(4): 764–778 4. Esposito V, Paolini S, Morace R (2006) Resection of a left insular cavernoma aided by a simple navigational tool. Technical note. Neurosurg Focus 21: e16 5. Grunert P, Muller-Forell W, Darabi K, Reisch R, Busert C, Hopf N, Perneczky A (1998) Basic principles and clinical applications of neuronavigation and intraoperative computed tomography. Comput Aided Surg 3: 166–173 6. Gumprecht HK, Widenka DC, Lumenta CB (1999) BrainLab VectorVision Neuronavigation System: technology and clinical experiences in 131 cases. Neurosurgery 44: 97–104 7. Hattingen E, Good C, Weidauer S, Herminghaus S, Raab P, Marquardt G, Raabe A, Seifert V, Zanella FE (2005) Brain surface reformatted images for fast and easy localization of perirolandic lesions. J Neurosurg 102: 302–310 8. Keles GE (2004) Intracranial neuronavigation with intraoperative magnetic resonance imaging. Curr Opin Neurol 17: 497–500 9. Kral T, Kurthen M, Schramm J, Urbach H, Meyer B (2006) Stimulation mapping via implanted grid electrodes prior to surgery for gliomas in highly eloquent cortex. Neurosurgery 58: ONS36–ONS43 10. Nimsky C, Fujita A, Ganslandt O, Von KB, Fahlbusch R (2004) Volumetric assessment of glioma removal by intraoperative highfield magnetic resonance imaging. Neurosurgery 55: 358–370 11. Roberts DW, Hartov A, Kennedy FE, Miga MI, Paulsen KD (1998) Intraoperative brain shift and deformation: a quantitative analysis of cortical displacement in 28 cases. Neurosurgery 43: 749–758 12. Rorden C, Brett M (2000) Stereotaxic display of brain lesions. Behav Neurol 12: 191–200 13. Schulze-Bonhage AH, Huppertz HJ, Comeau RM, Honegger JB, Spreer JM, Zentner JK (2002) Visualization of subdural strip and grid electrodes using curvilinear reformatting of 3D MR imaging data sets. Am J Neuroradiol 23: 400–403 14. Unsgaard G, Ommedal S, Muller T, Gronningsaeter A, Nagelhus Hernes TA (2002) Neuronavigation by intraoperative three-dimensional ultrasound: initial experience during brain tumor resection. Neurosurgery 50: 804–812 15. Wadley J, Dorward N, Kitchen N, Thomas D (1999) Pre-operative planning and intra-operative guidance in modern neurosurgery: a review of 300 cases. Annu R Coll Surg Engl 81: 217–225

Intraoperative localization of subcortical brain lesions

Comments MRI-based corticotopography was developed by the authors to localize mid-sized subcortical brain lesions without much technical efforts. The practical application showed that their method proved to be an easy tool for intraoperative guidance. In its use it more or less resembles the classical application of sulcal anatomy knowledge used as anatomical landmarks for intraoperative orientation. To develop the method further an automatic alignment of the photography of the sulcal pattern of the surgical site with the 3-D rendering of the MR dataset will be the next logical step. This method will then allow a patient registration without much user interaction, avoiding the common errors of fiducial-based patient registration procedures of standard navigation systems. Furthermore, the method also has the potential to increase the navigation registration accuracy when it would be used as additional information in a navigational setup, allowing to check for the navigational accuracy per se and also to correct for initial patient registration errors. Christopher Nimsky Erlangen, Germany Esposito et al. describe an original MRI-based corticotopography for defining the entry point on the cortex during surgery of subcortical

543 intraparenchymal lesions. This technique does not imply any significant financial investment due to the fact that the software used for brain extraction and 3d rendering (MRIcro) is freeware and the software to manipulate the intraoperative picture of the cortex (photo-editing software) is often standard, both running on a standard laptop. This MRIbased corticotopography seems fast and reliable in their 46 patients and so should be of particular interest to any department having no access to standard neuronavigation. However to develop such a technique when having access to a ‘‘standard’’ neuronavigation surprises me a little bit. Indeed, our experience with our neuronavigation system (Vectorvision, Brain Lab, Germany) is that such system is rapidly operational, less than 10 min, and particularly comfortable due to the transfer of neuronavigation data into the microscope. So based on their data and my experience the time-saving aspect of their technique still must be demonstrated. Besides, the brain-shift at the start of brain surgery is always minimal and not a problem for accessing subcortical lesion. Despite these controversial points, these authors must be congratulated for having developed a cost-effective MRJ based corticotopography for subcortical lesions localization in a time where cost-effectiveness is increasingly important. Christian Raftopoulos Universite Catholique de Louvain