The need to precisely localize targets within the brain and to refer them to important anatomical structures has occupied neurosurgeons since the early years of intracranial surgery. Craniometry, developed by neuroanatomists in the 19th century, was the first practical method of surgical navigation. It is still being used today as a crude but useful means to correlate the position of superficial brain anatomy with readily identifiable cranial hallmarks. Stereotaxis, first introduced by Horsley and Clarke in 1908 , represented the first big leap into deep brain localization. Advances in intracranial imaging and computer science enabled the evolution of stereotaxis, from a method that allowed the precise localization of a point in space in the human brain , to the capability of defining the whole contents of the head in a three-dimensional matrix [3-5]. This later development was crucial to intraoper-ative navigation. Computer reconstruction of images in different planes and three-dimensional rendering gave the neurosurgeon a display of anatomy that helps in the planning of surgical trajectories to deep-seated lesions and with the ability to "see” around the pathology to be treated. During the 1980s and early 1990s, several makers of stereotactic equipment developed computerized packages for volumetric stereotaxis in conjunction with ste-reotactic frames. Any frame reduces the working space available to the surgeon, particularly when the trajectory involves the lateral or posterior aspect of the skull. This limited the appeal of the newer computerized stereotactic systems, leading to the development of new approaches to neuronavigation.
There are two major paradigms for image-assisted neurosurgery: (1) real-time imaging, and (2) preoperative imaging. Real-time imaging is obviously the ideal approach to navigation, as it can correct for perioperative changes in the relative position and bulk of any pathology originated by the surgical procedure itself. Developments in this direction are mostly based on open MRI technology. Currently, these systems have disadvantages that compromise their widespread application: they are costly, require special sets of nonferromagnetic surgical instruments, and severely restrict the space available for optimal patient positioning and the neurosurgeon’s freedom to maneuver. Issues of safety resulting from prolonged exposure of medical personnel to radiofrequency energy are still being debated. Preoperative image-based systems have the advantage of relative simplicity, reasonable cost, and the potential for hassle-free handling.
Methods for frameless stereotactic neuronavigation include encoder articulated arms, ultrasonic digitizers, robots, and infrared active or passive flash/camera systems. Their description is beyond the scope of this topic. All these systems have minor problems in their integration to the regular operative suite environment.
Ideally, a neuronavigational system should allow the surgeon to work transparently, not being distracted by issues of instrument compatibility, detectors’ line of vision, etc. An ideal system would recognize immediately any instrument handed to the surgeon, transforming it into a pointing device, so that during its use, the computer display would continue to update its anatomical rendering from the surgeon’s point of view. This ideal is yet to be accomplished. The VectorVision system to be described here is a good step toward this goal.
BrainLAB VectorVision is an intraoperative, frameless, image-guided navigation system. The system integrates preoperative computed tomographic/ magnetic resonance imaging (CT/MRI) digital data, with real-time movement or position of surgical instruments. The navigation is based on reflective markers detected by infrared cameras .
The surgeon is able to point targets dynamically selected, in a frame-less environment.
The whole system in use at our institution is contained within a trolley that can be easily rolled in and out of the operating room. The trolley contains the cameras, the cameras’ controller, and the computer workstation. BrainLAB is now marketing a new version of the system (VectorVision 2). This new version, also based on the principle of easy transportation, has, aside from a sleek design, a touch-screen liquid crystal display that simplifies its use during surgery.
Two infrared-emitting cameras are mounted on a holder at a fixed distance of 100 cm from each other (Fig. 1). Infrared detectors are arranged around each camera, so that they act both as source and detector of the infrared light. The angulation of each camera is adjustable, both in the horizontal and vertical planes, so that the position of the trolley can be modified within a certain range (90 to 200 mm).
The system is based on an alpha 433 MHz microprocessor (Digital Technologies), running on Windows NT® 4.0 operating system and containing the VectorVision® software. A cameras’ controller represents the interface between the cameras and the computer (Fig. 1).
Reflective markers are plastic spheres with a glass-grain coating (Fig. 2). They reflect the infrared light emitted by the cameras. The detectors around the cameras read the reflection. These passive-reflective markers, first introduced by BrainLAB for navigation, are an important feature of the system. The lack of cables adds freedom for the surgeon during active navigation. These spheres can be mounted on any surgical instrument on a number of three-arm adapters (Fig. 2) with different geometries, which allow their separate recognition by the system. Thus, several instruments can work as pointers simultaneously. There is a basic pointer (Fig. 2), which is provided with two passive spheres and used for early registration. It may also be used for navigation throughout the procedure. All the reflective markers can be sterilized by gas or plasma. They have a limited lifetime of about ten procedures.
The Mayfield adapter has two components: a star-shaped tool to which three passive spheres are screwed and a fastener for the Mayfield headrest (Fig. 3).
Figure 1 The working station.
The star-shaped tool locks into the fastener very precisely. This tool is essential for the navigation procedure. It acts as a reference for the system, provided that the patient’s head does not move in relation to the Mayfield headrest during surgery. It has a calibration cone in its center, which allows for registration of normal instruments during surgery. The fastener may be applied to the Mayfield headrest under nonsterile conditions. The star-shaped tool may then be sterilely applied to the fastener for primary registration, and removed after registration is accomplished. The patient may be prepped and draped, and then the star-shaped tool reapplied to the fastener.
Figure 2 Reflective markers, skin markers, and Mayfield adapter.
For preoperative imaging, skin fiducials are attached with double-sided tape to the patient’s head. The fiducials have two components: a cone-shaped plastic, and three different metal markers: spherical for either MRI or CT scanning and hemispherical for registration in the OR (Fig. 2).
Only the Moller VM 900 microscope (J.D. Moller Optische Werke GmbH, Wedel, Germany) may be actively connected to the VectorVision system. The microscope is fitted with a special adapter containing reflective marker spheres. With this connection, the focus of the microscope is directed automatically to any place pointed to by one of the surgical instruments, freeing the neurosurgeon’s hands from this task .
For neuronavigation, the following steps are followed before actual surgery. 3.1 Preoperative Imaging
Preoperative imaging is obtained usually on the evening before surgery. The patient’s head is shaved in all places where skin markers are applied. At least three markers should be used, but precision is enhanced with up to five markers. The position of the markers is dictated by several considerations:
(1) Position of the lesion/target and the area of the planned skin incision;
(2) The planned position of the head relative to the cameras; and (3) The prospective position of the Mayfield pins. Obviously, no skin fiducial should be applied in the area where the Mayfield pins will be applied, as Mayfield fastening (which is done before registration) will substantially displace the skin around the pins. In addition, the headrest itself may interfere with visualization or registration of the fiducials. All the fiducials should be easily seen by the cameras and, consequently, should be clustered on the side of the head that will be uppermost.
Figure 3 Mayfield adapter. The clamp is attached to the Mayfield holder under nonsterile conditions. The star-shaped tool may then be screwed to the clamp with sterile gloves and detached if needed. Accuracy of reattachment is within 0.1 mm.
CT or MRI Scans
Scans are obtained, usually with contrast enhancement. A number of constrictions in imaging acquisition should be taken into consideration, as the BrainLAB software cannot recognize or process certain scanning formats.
The scan should be acquired without tilt. For better three-dimensional reconstruction, the slices should be as thin as possible and the scan should comprehend the entire head. The zoom factor should allow all skin markers to be visible. Scan parameters and the patient’s head should remain constant during the scan. Slice overlap should be avoided (VectorVision software will not recognize overlapping slices as separate). The matrix should be set to either 256 X 256 or 512 X 512.
It is recommended to place the patient’s head on a flat holder (such as the body holder), instead of the round-shaped head holder. This will enhance the usefulness of the three-dimensional view (the usual head holder will obscure the outline of the posterior-lateral aspects of the head in the three-dimensional reconstruction).
Imaging data are then transferred for processing to the VectorVision workstation using digital media or a communication network.
Preoperative Data Processing
The software provides a variety of options. Either CT, MRI, or both may be processed. An image-fusion module allows the surgeon to use both imaging modalities. In our experience, CT is sufficient for most navigations. The lesion and other areas of interest may be outlined so that they can be seen in 3D reconstruction. For the purpose of intraoperative navigation, the outlines may be switched off, so that true anatomical imaging can be used.
The Operating Room
There are several ways to organize the operating room. We use the cameras attached to the computer trolley, placing the trolley between the neurosur-geon and the anesthesiologist.
In any position, the trolley should be in the surgeon’s visual field, and the cameras about 1 to 2 m from the patient’s head. The two cameras can be moved freely as a unit during surgery. The cameras ought to have an unobstructed view of the star-shaped Mayfield adapter. When the operating microscope is to be used, the cameras are positioned as low as possible, so that the microscope will not interfere with their line of view.
Calibration of the cameras can take place any time, although it is desirable to perform it in the final position for surgery. The procedure, which takes only a few seconds, is done with a special tool having two reflective markers.
As the tool is moved in front of the cameras, the software reads several markers’ positions and displays the recommended working range.
The registration process includes digitizing the skin fiducials and the reference Mayfield headrest markers. This step is done after the Mayfield headrest is fixed to the head. The procedure can be done in a sterile or nonsterile manner (before or after the head is prepped and draped). We prefer the nonsterile method: the sterile star-shaped tool is handled with sterilized gloves, attached to the Mayfield headrest and angulated, so that the three reflective spheres are well within the field of view of both cameras. This is seen in an active window that opens in the computer display when the registration prompt is selected in the software. The skin fiducials are then digitized. The software gives a choice of automatic or manual registration. In the automatic mode, the software auto-detects the placement of each fiducial in the preoperative image and prompts the user to mark them on the patient’s head. This is done with a nonsterile pointer. When the pointer touches the fiducial, the software digitizes the position of the spheres in the pointer. A color-coded window in the computer display changes color when the digitization is done. When all the fiducials have been digitized, the computer display becomes active, and an accuracy figure is given (if the accuracy is less than 5 mm, the software will reject the digitization). The whole procedure is completed in less than a minute. From this moment on, neu-ronavigation is active. Anytime the pointer is brought into the cameras’ field of view, the computer display will show reconstructed axial, coronal, and sagittal images of the head centered at the pointer’s tip.
The skin fiducials are no longer needed once registration has been completed. After their removal, the patient’s head is prepped and draped as usual. The star-shaped tool is left above the drapes.
We use the split-screen mode during surgery, displaying coronal and sagittal views, as well as a choice between axial and three-dimensional view (Fig. 4). The three-dimensional view is useful to design the craniotomy flap. The other views are instrumental to navigation once the craniotomy is done. Obviously, changes in the geometry of the brain during surgery reduce the accuracy of the system. Common situations in which this may happen are brain shift after tumor debulking, brain shift after aspiration of cysts, aspiration or drainage of cerebrospinal fluid in hydrocephalic patients, and severe brain swelling/edema during surgery. All these situations may change brain configuration and make the system unreliable.
Figure 4 A typical display of the VectorVision system, with axial, coronal, and sagittal reconstructions during surgery, while the pointer is being introduced through a sylvian fissure approach to the target, a left basal ganglia cavernoma. The position of the internal capsule has been outlined. Surgery was done with the patient fully conscious.
We usually abort navigation under those circumstances. More unusual reasons for navigational failures are movement of the patient’s head relative to the star-shaped tool and electric power shutdown during surgery. In the first situation, which takes place when the head falls from the Mayfield headrest, navigation needs to be aborted, unless another patient’s head-contained reference system has been previously digitized . In the event of an electric power shutdown, the system has an option named ”recover” that can allow continuation of the procedure.
The VectorVision system is accurate enough for almost all surgical localization procedures. We have not encountered a single case in which the navigation failed to point to the target. The system advantages are the rapid setup requiring only a few extra minutes in any procedure, the easy handling of the pointer tools devoid of cables, and well-designed software tools.
The pointer tool can be used to plan the skin incision and the bone flap, and then to identify relevant anatomical structures. During tumor resection, it can help define tumor borders. Cavernous angiomas or small arte-riovenous malformations deep in the brain are excellent cases for neuro-navigation. With its three-dimensional and three-planar rendering, the VectorVision system is of great help in designing and maintaining the best approach to the target.
Attached markers can be used to guide catheter into ventricles. In this situation, it is advisable to use the transfrontal route. When using the trans-occipital route, the markers attached to the far end of the catheter can fall beyond the cameras’ field of view.
We have found neuronavigation ideally suited to assess the progress of surgery when dealing with large skull-base meningiomas. Because the remaining tumor attachment does not change its geometry during surgery, the pointing tool can show the exact position of the surgeon’s instruments at any stage.
Neuronavigational systems such as VectorVision modestly increase the cost and planning time of neurosurgical procedures by requiring additional imaging and hardware/software setups. They also require some time investment in learning to deal with a computerized system. Nonetheless, the added capabilities are worthwhile. The surgical procedure itself is better tailored to the pathology: smaller skin incisions and craniotomy flaps result in less postsurgical patient discomfort. This, in turn, may result in shorter hospi-talization time. The surgeon enhances his/her confidence when dealing with deep-seated intra-axial lesions, or large extra-axial tumors. There is no major change in the immediate operative environment and no bulky added hardware compromising the surgeon’s freedom.
The VectorVision system also has a spinal module useful in the planning and execution of fusion instrumentation, such as transpedicular screw placing and plating. Its discussion is beyond the scope of this article.