1 Collaborative Planning in Epilepsy Surgery

10.1055/b-0040-177282

1 Collaborative Planning in Epilepsy Surgery

Ignacio Delgado-Martínez, Rodrigo Rocamora, Gerardo Conesa Bertran, and Luis Serra

Abstract

The planning of stereotactic electrode implantation for epilepsy treatment involves the manipulation of complex patient multimodal information by a collaborative multidisciplinary team. This requires a workflow in which patient data and stereotactic electrode information are shared back and forth between epileptologists and neurosurgeons in order to progressively refine the trajectories. This process is divided into stages which cycle around until consensus is reached. The goal is to match best to the diagnostic requirements of the surgery with minimum risks for the procedure. The implantation is performed using guiding tools such as navigation or robots to ensure precision and efficiency. In the final stage, the implanted position of the electrodes is needed to accurately identify the electrophysiological sources of epileptogenic activity in the patient brain. Over the last years, many computer tools have been developed to help with the planning of implantation surgeries: processing of neuroimaging data, multimodal 3D visualization, automated planning algorithms, robot-assisted implantation, etc. By specializing in different tasks, these tools greatly facilitate the work of the clinical members involved in the implantation process. However, such specialization has resulted in a disconnection between the different tasks of the workflow and smooth and safe information flow. A different approach is the one provided by Collaborative Workflow Systems, such as the SYLVIUS platform developed by Hospital del Mar (Barcelona, Spain) which seeks to centralize the planning of workflow in a single platform in which the individual members of the team can perform their specific tasks in a precise and coordinated manner. This platform combines 2D and 3D data manipulation capabilities using a practical virtual reality interface.

1.1 Introduction

The goal of epilepsy surgery is to remove or disconnect the epileptogenic zone (EZ), that is, the area of the cortex that is necessary and sufficient for initiating seizures and, if not completely resected, reinitiates the pathological process. ¹ Procedures for treating epilepsy are not only limited to resective and palliative surgeries, minimally invasive procedures also provide good outcomes. ² Minimally invasive procedures typically consist of the local lesioning of the EZ by gamma radiation or by thermocoagulation with laser or radiofrequency pulses. The accurate determination of the EZ position and its propagation pathways is one of the fundamental factors that influence the outcome of the surgical procedure. The EZ location is deduced through a joint process of epileptologists and neurosurgeons, which ultimately leads to the definition of the therapeutic strategy. Such multidisciplinary involvement makes epilepsy surgery one of the most collaborative neurosurgical subspecialties.

This process starts already before the patient’s admission to the hospital, with several noninvasive examinations (▶Fig. 1.1). If the EZ or its relationship with functional zones is not adequately defined at this stage, invasive examinations are necessary. Invasive electroencephalographic recordings are often done by means of subdural electrodes, placed on the surface of the cortex through a craniotomy. Recordings from deep or otherwise inaccessible brain areas are obtained using long multicontact electrodes, which are surgically implanted following stereotactic trajectories. ³ This procedure is less aggressive and produces higher quality stereotactic EEG (SEEG) recordings than subdural grids and thus encourages the use of deep electrodes for the long-term monitoring of superficial areas as well. ⁴

**Fig. 1.1** Clinical workflow for planning of epilepsy surgery. Preadmission data (*blue*), preimplantation data (*red*), and postimplantation data (*green*) access the pipeline at the corresponding phase and flow downstream. At each phase, data are analyzed by the respective specialists producing a specific output, which will be used in the following phase. The decision at the end of the workflow is a therapeutic strategy for the patient.

The workflow of stereotactic electrode implantation can be divided into five phases (▶Fig. 1.1). During the epileptologist phase, the epileptology team establishes the initial hypothesis about the origin of the epileptic network according to the noninvasive examinations. Following this, in the neurosurgery phase, neurosurgeons create the electrode implantation plan based on the proposal of the epileptologist and the anatomy of the patient. Next, in the revision phase, the neurosurgical plan is reviewed by both teams to verify that the hypothetical location of the EZ is optimally covered by the exploration volume within the stereotactic trajectories. Only then, the plan is ready for the surgical implantation at the implantation phase. After surgery, the position of the contacts is determined during the validation phase and compared to the implantation plan. After this process of electrode implantation, SEEG recordings of the patient are obtained for 7 to 21 days at the epilepsy unit. The data collected during this period is used by epileptologists and neurosurgeons for deciding the final therapeutic plan.

1.2 Epileptology Phase: Establishing the Plan

The aim of the first phase of the epilepsy surgery workflow is to define the exploration volume based on the postulated location of the EZ and its relationship to cortical areas. During this phase, patients undergo diagnostic tests (▶Fig. 1.1), which includes a detailed history and examination, surface video EEG telemetry, and neuropsychological and psychiatric evaluation. The neuroimaging protocol typically involves structural MRI sequences such as three-dimensional (3D) T1-weighted (T1W) and T2-weighted (T2W) volumetric acquisition, fluid-attenuation inversion recovery (FLAIR) acquisition, and other inversion recovery acquisitions. Nuclear imaging modalities, such as fluorodeoxyglucose positron emission tomography (FDG-PET), ictal single photon emission computed tomography (SPECT), or subtraction ictal SPECT coregistered to MRI (SISCOM), allow the investigation of brain function and are an essential part of the evaluation of the epileptic network dynamics. Advanced neuroimaging protocols, mainly diffusion-weighted imaging (DWI) and functional MRI (fMRI), are important for the understanding of the propagation pathways and the relation to eloquent areas.

1.2.1 Computer Processing of Neuroimaging Data

Manual analysis of these data is challenging due to the considerable variability in the visual interpretation of the images. Several computerized methods can assist in this task. Image segmentation is commonly used for measuring and visualizing different brain structures (▶Fig. 1.2a). The purpose is to parcellate an image into a set of defined, homogeneous, and non-overlapping regions of similar attributes like intensity, depth, color, or texture. The segmentation result is either a new image made of voxels that replace intensities with masks for labels identifying each distinct region or a set of contours describing the region boundaries. ⁵ Other imaging diagnostic algorithms are specialized in detecting specific structural changes. ⁶ They extract measuring parameters by comparing signal intensity changes of standard imaging sequences (T1W, T2W, or FLAIR) between hemispheres or by computing topographical features of the gray matter, previously labeled with a segmentation algorithm. Those parameters can be correlated to certain structural abnormalities. For example, gyrification, sulcal depth measurement, and cortical thickness measurement are very sensitive to malformations of cortical development or to focal cortical dysplasias.

**Fig. 1.2** Segmentation techniques. **(a)** Cortical segmentation from a T1W sequence. **(b)** Vessel segmentation obtained from rotational angiography. **(c)** Electrode segmentation after postimplantation CT.

Neuroimaging processing algorithms are commonly coded with the SPM package of MATLAB (MathWorks, Inc.) or with the Insight Segmentation and Registration Toolkit (ITK) library in C++ (National Library of Medicine). Many published algorithms are available online in the Neuroimaging Informatics Tools and Resources Clearinghouse website (NITRC, https://www.nitrc.org/), which is supported by the NIH. Several open-source analysis suites for neuroimaging (e.g., FSL, ⁷ AFNI, ⁸ or FreeSurfer ⁹ ) contain numerous tools for registration, segmentation, and other more complex analyses of MRI data in one single package. Tractography maps can be obtained from DWI using specialized software, like Diffusion Toolkit, Camino, TORTOISE, MRTRIX, StarTrack, and others. ¹⁰ Other neuroimaging packages, such as 3D Slicer (Harvard Medical School), are specialized in multimodal visualization, but they also provide diverse tools for creating custom algorithms, which make them exceptionally versatile for neuroimaging. ¹¹

1.2.2 Planning Using Regions of Interest

With all the available information, raw as well as computer-processed, the epileptologists have the task of postulating the location of EZ and determining the regions of interest (ROIs) that will be monitored with intracerebral electrodes (▶Fig. 1.1). For temporal lobe epilepsies, the most frequent type of surgically treatable epilepsies, commonly targeted regions are the hippocampus, the amygdala, or the neocortex in the temporal or frontal lobe. Frequently explored regions in other epilepsy types are the parietal or occipital lobe. Often, the language cortex is considered as well so that speech function can be examined with SEEG recordings before a possible resective surgery. ¹² How the ROI definitions are communicated from the epileptologists to the neurosurgeons greatly varies in each institution. Frequently, the easiest and fastest alternative is to manually draw the ROIs on a Talairach grid (▶Fig. 1.3a, b). The obvious limitation of this approach is that it fails to capture the three-dimensionality of the volume to explore. 3D multimodal visualization systems, such as 3D Slicer, allow the creation of sketches directly on the real volumetric data of the patient ¹³ (▶Fig. 1.3c). The epileptologist can, therefore, describe the plan more accurately and transmit it to the neurosurgeon minimizing information loss.

**Fig. 1.3** Definition of ROIs for stereotactic implantation. **(a)** Proportional grid system proposed by Talairach and Tournoux. ¹⁴ **(b)** Implantation sketch over a Talairach grid. **(c)** Screenshot of a 3DSlicer session showing the proposed implantation plan. **(d)** Screenshot of an implantation plan sketched with the epileptologist module of SYLVIUS.

1.3 Neurosurgical Phase: Building the Plan

During this phase, the epileptologists’ implantation sketch is translated into stereotactic electrode trajectories. Planning electrode trajectories is a complex, time-consuming manual task performed by the neurosurgeons themselves. These trajectories are defined by an entry point and a target point, which indicate, respectively, the point where the electrode enters the skull and the point where it stops within the target structure to analyze. The neurosurgeon is required to integrate multimodal information to locate cortical regions, gray matter structures, and targets, and decide optimal trajectories for the sampling of the ROIs or targets, avoiding risky structures. Trajectories perpendicular to the skull are preferred because they are less prone to possible surgical tool slipping and bending during the drilling phase. Sulci should be avoided due to the vessel presence, arachnoid stretching, and bad recording quality. Moreover, the electrodes must be suitably spaced away from each other to avoid collisions inside the brain and oversampling in small regions. ¹⁵ Last but not least, the neurosurgeon has to ensure that the entry point is accessible by the implantation device during the procedure. Due to the high density of electrodes, each additional placing of an electrode often requires adjusting previously planned trajectories.

1.3.1 The Vascular Network during Trajectory Planning

One of the most important constraints to be considered during the neurosurgical planning is the relationship of the trajectory to the vascular network. This is crucial considering the inability to visualize cortical vessels during percutaneous twist drilling. Although the incidence rate of hemorrhage is under 0.2% per electrode, the consequences of an intracerebral bleeding are critical. ¹⁶ The visualization of the vascular anatomy may be done, invasively, by catheter 3D digital subtraction angiography (3D DSA) or, noninvasively, by gadolinium-enhanced MRI or CT angiography. Although no direct comparison between complication rates of stereotactic electrode implantation using either method has been published, DSA is generally preferred, as it provides increased anatomical detail. Besides, vascular automatic segmentation algorithms are available for both noninvasive and invasive angiographic modalities (▶Fig. 1.2b), but those optimized for catheter angiography are more robust and reliable, due to the higher quality of the images. ¹⁶

Multimodal 3D Planning

Surgical planning uses volumetric information stored in a stack of intensity-based images, usually acquired with CT and MRI scanners. Surgeons can view these images using specific 2D image viewers to build their own mental 3D model of the anatomy. There is a variety of software available today to render CT or MRI volumetric models of the patient in real time to allow surgeons to better understand brain morphology. One popular example is Osirix (Pixmeo, Switzerland). However, planning electrode trajectories generally requires, in addition, to understand the relationship between electrode trajectories and brain structures, which is possible only with special software.

Planning of navigation or robot-assisted implantation procedures (see Sections 1.5.2 [Frameless Implantation Approaches] and 1.5.3 [Robot-Assisted Implantation Approaches]) is usually done using the specific native software provided by the manufacturer (▶Fig. 1.4). Since this software needs to be regulated by the health authorities to ensure the safety of the procedure, strong restrictions apply to what the user can do. Navigation planning systems are designed for a broad range of neurosurgical purposes, being, thus, more versatile and flexible than the native robot software. Nevertheless, these systems are limited, for example, in the type of the neuroimaging data that can be imported. These and other limitations inherent in native software (either in robots or navigation systems) are addressed by recent software tools. Of note is EpiNav (University College of London, UK), which supports a great variety of multimodal neuroimaging data and offers tools for multiple electrode trajectory planning and automated trajectory planning ¹⁷ (▶Fig. 1.5a). Unfortunately, being still a research tool, EpiNav is not easily available outside their research environment and does not have yet the regulatory certification for medical use.