Imaging and Fiducials

One or more volumes of image data (e.g., MRI, CT, functional magnetic resonance imaging [fMRI], positron emission tomography [PET]) are acquired containing most of the patient’s head (and, of course, the brain tumor). These data must be spatially accurate, and at least one of the volumes must contain some surface reference marks or features that can be accessed at surgery. Typically, these are adhesive markers (called fiducials) that are applied to the scalp in a wide distribution, but other markers can include anatomic features, skull-implanted fiducials, and feature contours. Scalp fiducials should be multiple and widely spaced, should not appear along a straight line, and should be located over relatively immobile scalp. With skull-implanted fiducials, the overall accuracy of navigation may be submillimetric and may match or exceed that obtained with stereotactic frames.^16,17

Registration

The most common means of correlating (or “registering”) image data with the physical space of the patient’s head is called paired points.^18,19 At surgery, the reference points are identified on the images and are touched with a pointing device. When surfaces are used, the physical surface is matched, or registered, to that of the radiographic surface, either by touching multiple random points on the surface (“cloud of points”) or by scanning the surface with laser beams.²⁰

Pointing Devices

A variety of three-dimensional digitizers have been used to allow the navigation computer to locate the surgical pointing device in space. Historically, these have included mechanical arms with multiple articulations (both analog and digital), and ultrasonic, machine vision, and various magnetic devices.^21–30 Today, most systems use active or passive (i.e., reflective) infrared markers on the pointing device, with the position determined by stereoscopic solid-state cameras that locate the markers trigonometrically.^31–33 When the geometry of the markers and the pointing device are known to the computer, it can locate the tip and axis of the pointing device in the operating room. One disadvantage of this method is that it requires that line of sight be maintained between the probe markers and cameras. This can, at times, be logistically difficult, particularly when an operating microscope is to be used. Although a microscope can be adapted or designed to serve as such a pointing device, other technologies such as electromagnetic digitizers may be better suited for such applications.^23,28,34

Display

When the registration process is completed, the computer can exhibit the location and orientation of the wand on the image data and can even be used to guide the surgeon to a preselected target along a prescribed trajectory. Common displays include one that shows a set of coronal, axial, and sagittal planes that converge at the point of interest and one that shows planes that are steered by the pointing device, including along the axis of the pointer or perpendicular to the axis.^21,22 Three-dimensional presentations may also be made when using the system for guidance along a prescribed trajectory.

Other sets of image volumes may be “fused” with the primary data set. This multimodality integration allows for more versatile use of navigation (see later) or use of images in which the scalp references are not visible. The process may be manual, automated, or a combination of methods.

These images are typically displayed a few feet from the surgeon on a large flat panel display, or the older cathode ray tube technology may be used. Devices to improve the presentation of navigation data include various head-mounted displays, including light-emitting diode and laser technologies, but these have not been adopted widely.³⁵

Patient Head Movement

Fundamental to the facile use of navigation in the operating room is the ability to compensate for patient or detector (e.g., camera) movement. Although this may be accomplished by indirect mechanical linkage of the detector to the patient’s head, it is more often done by use of a set of reference markers that are linked to the patient’s skull—usually attached to the patient’s head holder. These dynamic reference frames (DRFs) are essential to the use of today’s infrared systems. It is, however, imperative that the DRF be securely affixed so that its relationship to the head cannot be disturbed after the registration procedure.

Brain Movement

Perhaps the greatest limitation to use of surgical navigation is movement of the brain during surgery compared with the preoperative state when the images were obtained. Gross movements of the brain occur after the dura is violated owing to loss of cerebrospinal fluid and are most prominent over the convexity and poles.^18,36 Significant brain “shifting” is a problem that may occur during biopsy as well as during craniotomy for tumor. Fortunately, both these lobar displacements, as well as local distortions due to surgery, can usually be managed with some surgical foresight and are discussed later.³⁷ In certain cases, however, intraoperative imaging may be required to compensate fully for these movements.

Minimal and Optimal Access Craniotomies

Although it was once deemed that “the only good craniotomy is a large craniotomy,” image-guided stereotactic techniques and, in particular, surgical navigation have led to smaller, strategically placed and sized craniotomies. This change has occurred mainly because earlier surgery was largely exploratory, and targeting was often based on the appearance and “feel” of the brain rather than on the precise guidance offered by surgical navigation.

Minimal access craniotomies may have several advantages, including reduced length of surgery, lower incidence of wound infections, and shorter length of hospital stay.^10,38 The minimum size of a craniotomy is, in part, dependant on the size and depth of the lesion as well as on surgical instrumentation. For intraparenchymal lesions at the cortical surface, the craniotomy generally should be large enough to encompass the extent of presentation of the tumor on the surface. For deeper lesions, the craniotomy may not need to be as large as if the lesion presented at the surface because the skull opening can be considered the apex of a working cone extending down to the tumor (Fig. 118-1). Of course, the opening must be large enough for the surgical instruments to fit, as well as for proper illumination and visualization of the region of work. Endoscopic procedures may be performed through very small openings (e.g., bur holes), whereas most microsurgical procedures require a minimum of 2- to 3-cm craniotomies. Extra-axial lesions such as meningiomas may require large craniotomies, but these can be optimized to account for dural tails, surface and draining veins, and intended extent of resection.^39,40

FIGURE 118-1 A, Even large lesions may be accessed through a minimal access craniotomy when at sufficient depth. B, A similar-sized lesion near the skull, however, requires a larger, optimal access craniotomy.

Relationships to Critical Brain

Another important preoperative role of surgical navigation for brain tumors is to reduce the risk for neurological morbidity by allowing the surgeon to determine the relationships of the lesion and surgical approach to nearby critical brain and structures (e.g., vascular anatomy). Often, this information may be gleaned from anatomy alone, and navigation allows for unique views of the cortical surface that may resolve ambiguities of surface anatomy compared with reliance on traditional axial, coronal, and sagittal presentations. Further, visualization of critical surface or draining veins may be facilitated using these systems.

Most contemporary navigation systems allow for fusion of one or more image data sets (see the discussion of multimodality integration, later) for planning and navigation. Functional image data such as PET, fMRI, and magnetoencephalography may augment anatomic data when superimposed on high-resolution MRI (Fig. 118-2). Perhaps the greatest advance in navigation in recent years is the ability to incorporate diffusion tensor imaging (DTI) fiber tracking (Fig. 118-3) into the image data set.⁴¹ Although navigable anatomy was largely limited to cortical, periventricular, and lesion features, DTI fiber tracking has led to the era of subcortical navigation, particularly when used with subcortical stimulation of fiber tracks.

FIGURE 118-2 Coregistered functional magnetic resonance imaging (MRI) shows motor activity (magenta), tumor (green), and corticospinal tracts (yellow) superimposed on high-resolution volume MRI in navigation system. Corticospinal tracts are derived from diffusion tensor MRI using a fiber-tracking algorithm.

FIGURE 118-3 Display of white matter tracts derived from diffusion tensor imaging fiber tracking. Orange represents optic radiations, purple shows corona radiata, and pink shows tumor.

Guidance to Subcortical Lesions

Because there are few, if any, landmarks within the substance of the brain, unambiguous guidance to subcortical tumors is another important function of surgical navigation. Contemporary systems manage this guidance function in one of several ways—usually by projection of the axis of the pointing device onto traditional axial, coronal, and sagittal displays or orthogonal views along and perpendicular to the axis of the pointing device.⁴² My experience has been that the latter is most useful because it shows all structures that will be encountered en route to the tumor, and thus the surgeon can “step down,” millimeter by millimeter, to the tumor with a view perpendicular to the pointing device or intended trajectory (Fig. 118-4).

FIGURE 118-4 Multiple potential trajectories to this hypothalamic tumor are shown. Top right and bottom left panels contain the purple trajectory; bottom right panel is perpendicular to the trajectory.

Assistance with Resection Control

Assisting with resection control is perhaps the most misunderstood and underappreciated function of surgical navigation for intra-axial tumors. Some respected neurosurgeons have asserted that navigation is of no use in aiding determination of how much of the tumor has been removed, compared with the preoperative image data set, because of local tissue deformations caused by the procedure. Although local and lobar brain movements do occur after surgery, adjustment of the technique of tumor resection can allow navigation to be an important aid in resection control.^38,43,44

On opening the dura, cerebrospinal fluid will begin to drain from the wound, resulting in a gradual drift of the brain as it deflates from loss of ventricular pressure and volume. With standard head positioning, this brain shifting can become pronounced during the course of a tumor resection and could lead to erroneous navigation using preoperatively acquired images. Fortunately, this problem can usually be managed, but not eliminated, by taking advantage of the fact that the shift is generally straight down toward the center of the Earth.⁴⁵ By orienting the patient’s head so that the surgical trajectory is vertical, the surgeon need only compensate for brain shift in one direction (i.e., the brain and tumor are lower than expected) rather than for a complex three-dimensional slide that may occur when operating from a different direction. Also, minimizing the use of diuretics and compensating for volume loss by limiting or reversing hyperventilation may be useful strategies. When only part of the resection involves critical brain, the surgeon should work on that area first, while shift is minimal.

Traditionally, surgeons are taught to remove intra-axial primary tumors in a centripetal (i.e., from inside-out) fashion—resection continues until “normal” brain is encountered. This approach, however, is almost guaranteed to lead to erroneous information from a navigation regarding extent of resection when only preoperative imaging is concerned. Such piecemeal resection usually leads to a collapse of the tumor, such that its maximal dimensions are less than its preoperative state. Therefore, when performing a centripetal resection, reliance on preoperative imaging will usually overestimate the amount of residual tumor and risk that the surgeon will stray into normal and, perhaps, critical surrounding brain. Again, modification of surgical technique can minimize this problem and allow navigation to be useful in this setting. Foremost among these strategies is to remove the tumor, as much as is practical, in an en bloc technique, similar to the traditional method of removing an arteriovenous malformation. The surgeon identifies the radiographic brain-tumor interface and microsurgically defines it until the tumor is isolated and removed, ideally in one piece. Although not always practical, usually large portions of the tumor can be resected in this fashion. Also, the surgeon should avoid puncturing any cystic components or entering the ventricles until all critical areas of the tumor boundary have been surgically defined.

When the surgical plan suggests that these strategies will fail (e.g., a large cyst is overlying surgical access to the tumor), the surgeon may use navigational guidance to place several markers around the borders of the tumor.^46,47 This “fence-posting” or “picket-fencing” may be time-consuming, however, and does not guarantee confinement or resection to the desired area.

Finally, I have found that placing large cotton balls in the resection cavity will usually expand the cavity to, but not more than, the preoperative dimensions. The surgeon can then explore the cotton ball–brain interface with the navigation system and visually interrogate areas where there appears to be residual tumor on the navigational images.

Therefore, a surgeon with a willingness to modify his or her technique of tumor resection can make use of surgical navigation as an aid to resection control. A summary of these methods is provided in Table 118-1.

TABLE 118-1 Methods to Minimize “Brain Shift” and Local Tissue Deformation for Tumor Resection

Target Selection and Procedure

I prefer to select the center of the most radiographically abnormal region of the tumor as the target for biopsy.⁵² This approach applies to both enhancing and nonenhancing lesions. The surgeon may choose to augment targeting with PET, magnetic resonance spectroscopy (MRS), MRI cerebral blood volume, or other means so that the most abnormal area of the brain is targeted for biopsy and the risk for underestimating tumor grade because of sampling error is minimized.⁵³

After a minimal hair shave and skin preparation, a small incision is made along the biopsy trajectory. I usually create a twist drill hole, whereas some surgeons prefer a bur hole. After dural perforation, a side-cutting biopsy instrument is advanced to the target site, and the outer sheath is left in place when extracting specimens. This allows for drainage of any bleeding and also impales the tumor on the instrument, limiting any effect of brain shift or local tissue deformation (e.g., tumor cyst fluid drainage). At least two samples are obtained, and they are ideally split longitudinally for frozen section or smear. Intraoperative pathology review is important because the first specimen is diagnostic in only about two thirds of cases. When the specimen demonstrates only necrosis, deeper or shallower biopsy samples almost always yield the diagnostic tissue of the enhancing ring. Targeting the ring risks missing the lesion entirely or results in sampling potentially critical brain adjacent to the tumor. The surgeon should bear in mind that if a vertical trajectory of the biopsy instrument is used, the brain may have shifted down and away from the surgeon, and a greater than predicted depth may be necessary to access a specified target.

Bleeding during brain biopsy can often be managed without resorting to emergency craniotomy.⁵⁴ When using a side-cutting biopsy instrument, the outer portion should be left in place to allow blood a route of exit from the site of bleeding. Irrigation and mechanical obturation of the cannula are essential to keep the device patent. Head elevation and blood pressure control may also be of benefit; the former is particularly important for venous bleeding. I have found that arterial bleeding may often be controlled by injecting a small net volume (about 0.5 mL) of thrombin (delivered by loading the inner biopsy apparatus with thrombin and replacing it in the outer sheath) and leaving the device occluded for 30 seconds before reopening. It is, again, essential to ensure that the bleeding has stopped as assessed by a patent cannula.

When adequate tissue is confirmed, the instrument is withdrawn, a few sutures are placed, the scalp is cleansed, and a dressing is applied. The patient is awakened and transferred for observation. A CT scan of the brain is obtained about 2 hours later to assess for occult intracranial blood. If the scan shows no more than 1 cm of blood (and is otherwise unchanged from baseline), undue bleeding did not occur at surgery, and the patient is neurologically unchanged, the patient may be safely discharged home to proper supervision.⁵⁵

Related Procedures

Tumor cyst drainage, either episodic at surgery or by placement of an Ommaya reservoir, may be useful in the management of these disorders.^56,57 Navigational endoscopes may allow for fenestration of tumor cysts or biopsy of intraventricular lesions.^26,58 Conversely, radioactive liquids (e.g., P32, colloidal gold) may be instilled into cysts with the aim of reducing fluid production. With appropriate dosimetry software, navigation systems may be used to place temporary or permanent radioactive seeds for brachytherapy.^59–61

Multimodality Imaging

The power and flexibility of these systems have been recently enhanced by incorporation of multimodality integration, whereby image data from different devices or times are fused. Fusions may be within modalities (e.g., different types of MRI) or across platforms (e.g., CT and PET). Instead of repeating the MRI study, fusion may help lower cost by allowing previously acquired nonstereotactic images such as MRI to be fused to a low-cost stereotactic scan such as CT without contrast. Fusion of MRI with CT may also help detect spatial inaccuracies in the former.^83,84 The visualization of certain low-grade tumors may be enhanced by fusing color-encoded fluid-attenuated inversion recovery (FLAIR) images with high-resolution volume MRI (Fig. 118-5). PET, cerebral blood volume, or MRS maps may be fused with a stereotactic study to identify the ideal target for brain biopsy.

FIGURE 118-5 Colorized low-resolution fluid-attenuated inversion recovery MRI superimposed on high-resolution volume MRI helps to better visually define this low-grade, nonenhancing glioma.

Economics of Navigation

Although sparse, most reports indicate that these devices are cost-effective, may reduce surgical morbidity, and enhance outcome.^85,86 These studies, however, are potentially troubled by patient selection or they compare patients from different time periods. Randomized trials to evaluate this better are unlikely in the foreseeable future.

Future Applications

During the next decade, it is likely that there will be an expanded application of intraoperative MRI with integrated navigation into the neuro-oncology operating room. Strategic implantation of catheters using navigation for convection-enhanced or local delivery of new therapeutic agents, such as immunotoxins, gene therapies, and new receptor-targeted drugs, will likely be an important part of future brain tumor treatment. Although robotic navigation is not new,^87,88 it may become more widely accepted as the need for these procedures increases.

Another potential advance in navigation for brain tumor is the use of magnetic field manipulation to move catheters and small delivery vessels through the brain. This “nonlinear” stereotaxy may allow biopsy from multiple sites to minimize sampling error through a single twist drill hole, and even flexible endoscopic resection of intra-axial tumors.