Chapter 17 Stereotactic Irradiation

Linear Accelerator and Gamma Knife

Derek R. McHaffie, Deepak Khuntia, John H. Suh, Wolfgang Tomé, Minesh P. Mehta

Radiosurgery was first described in 1951 by Lars Leksell,¹ a neurosurgeon at the Karolinska Institute in Stockholm, Sweden. The term radiosurgery was selected because of the similarity of this technique to stereotactic neurosurgery. The initial work by Leksell first involved the treatment of patients with functional disorders and benign conditions such as chronic pain and arteriovenous malformations but later included benign and malignant tumors. The first patient was treated with the Gamma Knife in 1968.² The limited availability, skepticism regarding the technology, and prohibitive costs initially limited the spread of Gamma Knife technology. By the 1980s, linear accelerator-based radiosurgery had evolved and further renewed the interest in stereotactic radiosurgery.

Because of hardware and software improvements, the use of radiosurgery has proliferated and several hundreds of thousands of patients have now been treated. In this chapter, we discuss the physics and radiobiologic principles of radiosurgery, as well as the clinical application of this technology. Particle beam therapy is not the focus of this chapter.

Basic Principles Of Radiosurgery

Radiosurgery differs significantly from conventional fractionated radiotherapy. By definition, radiosurgery implies a single treatment. With conventional fractionation regimens, some normal brain tissue adjacent to a target receives a nearly full dose of radiation. Normal brain parenchyma is relatively unforgiving in regard to late toxicity and, therefore, the fact that radiosurgery can treat with high-dose gradients adjacent to a nonmobile target makes the use of radiosurgery in the brain attractive. The use of a very large number of beams ensures that no individual beam contributes significantly to the cumulative dose, so that the amount of radiation delivered to normal tissues in the beams’ paths is minimized. This geometry provides physical dose distribution advantages for relatively small targets, usually less than 4 cm in diameter. Targets larger than 4 cm result in an unacceptable increase in dose to adjacent normal brain tissue.

Radiosurgery can be performed using various devices, including the multicobalt unit known commercially as the Gamma Knife (GK), particle beam devices, or modified linear accelerators. With technologic advances in software and hardware, a clear superiority of one technology over another has disappeared and mature clinical results demonstrate no outcome differences based on technology.³ A comparison of some of the features of the Gamma Knife and linear accelerator techniques is presented in Table 17-1. Because the linear accelerator-based units can also treat non-radiosurgery patients during “downtimes,” cost and volume considerations are key features that determine equipment selection.³

TABLE 17-1 Comparing Gamma Knife and Linear Accelerator Radiosurgery Characteristics

	Gamma Knife	Linac Radiosurgery
Clinical experience	Over three decades	Over two decades
Accuracy	Submillimeter	Submillimeter
QA	Fewer QA checks	More QA checks
Machine use	Dedicated machine	Usually not a dedicated machine
Functional disorders	Longer experience	Shorter experience
Can be used for extracranial radiosurgery	No	Yes
Price	High; must replace source every 5-7 years	Less expensive; no source replacement
Tumor location	Difficult to radiate peripheral lesions	Can treat peripheral lesions
Fractionated treatments	Not practical	Can treat with fractionated regimens
Treatment time	About equal	About equal

QA, quality assurance.

Radiosurgical and neurosurgical approaches are often complementary, but there are key differences. Radiosurgery does not require a craniotomy or general anesthesia and is performed on an outpatient basis. In a cost-effectiveness/cost-utility analysis by Mehta and colleagues ⁴ in the 1990s, the authors showed that the average cost per week of survival for a single brain metastasis was $310 for radiotherapy, $524 for resection plus irradiation, and only $270 for radiosurgery plus whole-brain irradiation. With the advent of shorter hospital stays following neurosurgical procedures, these numbers have probably changed somewhat. Other advantages include lower postoperative risks for bleeding or infection and rapid recovery times. Patients who are working before treatment often can return to work within 1 or 2 days. More importantly, neurosurgically inaccessible lesions and patients deemed unfit for surgery or anesthesia can often be treated with radiosurgery.

The clinical applications of radiosurgery have grown substantially over the past decade. There are multiple disease processes where the use of stereotactic radiosurgery has a well-established and defined role. Current indications, discussed later in greater detail, include arteriovenous malformations, benign brain tumors, malignant brain tumors, and functional disorders (Table 17-2). With improved ability to localize targets throughout the body, the principles of radiosurgery have expanded to include extracranial applications, with emerging roles in the management of lung, hepatic, and spinal tumors, among others. Some concepts of the radiosurgery paradigm have been incorporated into a strategy employing fractionated treatments; this approach, referred to as fractionated stereotactic radiotherapy (FSRT), is applied to both intracranial and extracranial lesions.

TABLE 17-2 Clinical Applications of Radiosurgery

Radiosurgery is an important tool available to the neurosurgeon and radiation oncologist. Proper implementation of the treatment requires collaboration among neurosurgeons, radiation oncologists, and medical physicists. This allows for thoughtful coordination of care, improved quality assurance, reduction in practice variation, a strategic marketing advantage, and improved patient satisfaction. Each specialty brings unique skills that are required for a successful radiosurgery program.⁵

Physics Of Radiosurgery

Stereotactic radiosurgery (SRS) involves the use of numerous beamlets of radiation aimed precisely at an immobilized target to deliver a single session of high-dose radiation. Though no single beamlet carries significant weight, a large dose is deposited at the intersection of these beamlets with a steep dose fall-off outside the target. As tumor size increases, this fall-off becomes shallower ⁶; typically, radiosurgery becomes prohibitive at sizes in excess of 4 to 5 cm.

Physics Fundamentals

Multicobalt units containing cobalt-60 (⁶⁰Co) sources include devices such as the Gamma Knife and the Rotating Gamma System (developed in China) (Figs. 17-1 and 17-2). ⁶⁰Co decays with a half-life of 5.26 years, requiring that sources be replaced every 5 to 7 years. For new sources, total activity is approximately 6000 Ci with a dose rate of 400 cGy/minute. Radioactive decay releases two gamma rays with an average energy of approximately 1.25 MeV. A gamma ray and an x ray cannot be discriminated and differ only in source of origin. Gamma rays result from radioactive decay, whereas x rays are generated by accelerating electrons (in a linear accelerator) and colliding them into a tungsten target.

Figure 17-1 A, Gamma Knife model C unit. The Gamma Knife unit is the most common gamma ray–based radiosurgery unit used in the world. B, The model C unit is equipped with an automatic positioning system (APS) that reduces the need for manual position adjustments during treatment. Pictured is the 14-mm collimator helmet. C, Perfexion, the newest Gamma Knife system, incorporates automated collimator changes and expands the treatment volume to include the cervical spine and head and neck regions.

Figure 17-2 Rotating Gamma System. This system was developed in China, but the device is commercially used in a few centers in the United States.

Courtesy Tomasz Helenowski, M.D., Rotating Gamma System Institute, Gurnee, Illinois.

Protons are charged particles that have been adopted in a few centers for use in radiosurgery. One of the advantages of protons is that they deposit most of their energy over a finite distance, with very little exit dose. This narrow region of energy deposition is known as the Bragg peak and it may allow for a reduction in integral dose (total dose to the patient).

Components of a Radiosurgery Unit

Both multicobalt and linear accelerator units have at their heart a source of radiation (cobalt for the Gamma Knife [GK] and the linear accelerator for linac-based units), a couch, a stereotactic helmet or frame system to ensure precise localization and immobilization, and sophisticated treatment planning software. When properly calibrated, an isocenter alignment to better than 1 mm can be achieved with either the GK or the linac radiosurgery system, although it is important to recognize that image resolution, which depends on the slice thickness, is usually of a cruder magnitude.⁷

The use of the stereotactic frame ensures reproducible patient position and serves as the platform to which various coordinate localization devices can be attached. Generally, the frames are applied under a local anesthetic and sedation with multiple pins that penetrate the skin and fix to the outer table of the skull. Coordinate systems used for radiosurgery are based on Cartesian coordinates and allow for the precise definition of the target relative to the origin of the stereotactic localization system one is using. This permits precise geometric definition for the placement of various isocenters (or “shots”) and the correct beam geometry. The coordinate system is established in the treatment planning system and is used during treatment planning to define each of the shots or groups of arcs used to deliver focused radiation to the target.

Pretreatment Quality Assurance

Because GK radiosurgery uses live sources, radiation exposure and safety are important issues. Exposure maps of the entire room and surrounding environment are therefore necessary. The central dogmas of radiation safety—namely, time, distance, and shielding—are paramount, especially when live sources are used. Principles of ALARA (as low as reasonably achievable) need to be implemented. Typical “action levels” include level I with 10% exposure (125 mR/quarter) and level II with 30% exposure (375 mR/quarter) of the maximum allowable. Obtaining these action levels will result in consequences, which vary depending on the rules set by the specific institution’s radiation safety department.

In regard to actual operation, only trained oncologists or therapists can operate radiosurgery units. All operators must be in the vicinity of the control area during the treatment. The operator will ensure that all required safety checks have been performed and that the written directive has been completed. In addition to their involvement in calibration and treatment planning, physicists play a major role in ensuring that radiation safety measures are in concordance with regulatory guidelines.

An in-depth description of the various linac radiosurgery systems is beyond the scope of this chapter, and the interested reader is referred to Task Group Report 54 of the American Association of Physicists in Medicine.⁷ In general, linac radiosurgery systems can be classified into three groups. First, there are couch-mounted systems that use the treatment lasers or implanted fiducial markers as their target verification system; these systems are probably the least accurate because they have an isocenter stability of at best ±2 mm for any couch and gantry angle. Next, there are the so-called floor stand–mounted systems that connect to the treatment couch and allow the user to correct for inaccuracies in the rotation of the couch during treatment. However, these systems do not allow for corrections of inaccuracies in gantry rotation and, therefore, such systems have at best an isocenter stability of ±1 mm for any couch and gantry angle. The most accurate linac radiosurgery system uses a floor stand with a high-precision bearing that is indexed to the linac but does not connect to the treatment couch. In this system, the floor stand is decoupled from the linac treatment couch; this allows for a high accuracy of couch rotation, on the order of a few hundredths of a millimeter. Furthermore, this system has a separate arm rotating around a high-precision bearing that holds the collimator, which is coupled to the linac gantry using a gimbal bearing. Because all translational and rotational degrees of freedom are decoupled from the linac, this system can be calibrated to have an isocenter stability that is equal to that of a GK of ±0.25 mm for any gantry and couch angle.⁸

Regardless of which radiosurgery system used, a comprehensive pretreatment quality assurance (QA) program should be implemented. A typical program consists of three parts. First, verification that the frame has not slipped by using a depth helmet to measure the distance from the patient’s skull to the helmet by means of various predefined access points. A measurement is taken before the patient is imaged, after the patient is imaged, and before the patient is set up on the floor stand. Only if all measurements are within 1 mm does one proceed with the treatment. Second, a pretreatment isocenter verification film is obtained. In this test, the coordinates of the first isocenter are set up on the floor stand and a QA phantom independently. This test film serves two purposes: it checks that the isocenter has been correctly set up on the floor stand, and it allows confirmation of the integrity of the radiosurgery system. Third, a thorough, independent pretreatment check is carried out by the attending physician, including verification of the isocenter coordinates, the collimator setting, the collimator angle, the cone size, the system interlocks, and a final check verifying patient positioning.

Radiobiologic Considerations

Radiobiologically, the use of a single fraction to treat a small volume differs significantly from conventional fractionated external beam irradiation (EBRT) procedures. This poses unique challenges in that the increased biologic effect on normal tissues is more pronounced than it is on tumors ⁹ (Fig. 17-3). The most important factor influencing the risk of developing late side effects is fraction size. Therefore single-fraction radiosurgery needs to be exquisitely accurate and must minimize the dose to normal tissue.

Figure 17-3 Representation of radiosurgery versus radiotherapy dose for both early- and late-responding tissues.

From Mehta MP: The physical, biologic, and clinical basis of radiosurgery, Curr Probl Cancer 19(5):265-329, 1995.

Another potential disadvantage of using single-fraction treatment is the inability to exploit the temporal effects of cell cycle distribution. Because cells are thought to be most sensitive in the G2 and M phases of the cell cycle, fractionating radiation treatment allows redistribution of cells into the radiosensitive phases, which may improve cell kill. Hypoxia is another important variable to recognize. Malignant tumors are usually partially hypoxic, and fractionated treatments take advantage of the reoxygenation of the tumor that occurs between fractions. The use of single-fraction SRS does not exploit cell redistribution and reoxygenation for therapeutic gain.

Despite these theoretical disadvantages of single-fraction radiation, efficacy with minimal toxicity has been shown in clinical trials in a multitude of disease sites. Radiosurgery, by treating in a single-fraction regimen, minimizes the deleterious effects of repopulation (tumor regrowth between fractions). Further, the high doses of radiation could potentially overcome the disadvantages of single-fraction delivery. Recent observations suggest the DNA-centric model of classical radiobiology is more complex than initial experiments appreciated, including effects on membrane-bound signaling pathways.¹⁰ It appears that radiation effects on the tumor microvasculature are also an important component of tumor response and control.¹¹ Higher ablative doses of radiation have been shown to induce marked endothelial apoptosis with substantial downstream effects on the competency of tumor microvasculature.¹¹ At doses of more than 8 Gy per fraction, the linear quadratic (LQ) model does not accurately predict cell kill or toxicity.¹² This poses significant clinical challenges in developing dose-volume normal tissue constraints for SRS and stereotactic body radiotherapy (SBRT), as the LQ model is often used to account for differences in dose per fraction when extrapolating from toxicity analyses obtained using conventional regimens.

Treatment Planning

Treatment planning methods vary based on the software, but ultimately the treatment plans are similar. The GK system contains 201 ⁶⁰Co sources that are channeled through a tungsten collimator helmet generating narrow radiation pencil-beams to coincide within 0.1 mm of each other at the geometric center of the helmet. Multiple circular collimators are used to generate “shots” to fill a particular volume using one or multiple isocenters. Doses are typically prescribed to the 50% isodose line. There are four helmets with varying collimator sizes of 4, 8, 14, and 18 mm used in the GK system. The latest GK unit, Perfexion, provides automated change of collimator helmets, reducing treatment times when multiple collimator sizes are employed.

Multiple linac-based radiosurgical systems are commercially available, including Novalis BrainLAB (BrainLAB AG, Feldkirchen, Germany), Radionics XKnife (Integra, Plainsboro, New Jersey), TomoTherapy (TomoTherapy, Inc., Madison, Wisconsin), CyberKnife (Accuray, Inc., Sunnyvale, California), and others. With linac-based radiosurgery, optimal distributions can be obtained by using a micro-multileaf collimator or circular collimators that “shape” the beam. These devices allow for fractionation, more flexibility with beam shaping, and also modulation of beam intensity. Linac-based devices are also more readily adaptable for extracranial radiosurgery or fractionated delivery.

Both GK and conventional linac-based SRS treatment planning procedures begin with the placement of the stereotactic head frame on the patient. The head frame is used for precise localization and rigid immobilization to guarantee that the patient’s position from the time of imaging to the time of treatment does not change. Before the stereotactic CT is acquired, a stereotactic localizer ring is attached to the head frame. For linac radiosurgery, the most commonly used independent stereotactic localization system is the Brown-Roberts-Wells (BRW) system (Fig. 17-4). The BRW localizer has an inner diameter of 29 cm and a height of 16 cm and consists of nine carbon fiber rods that are arranged into three N-shaped structures between two aluminum rings, which are placed at 120-degree angles around the rings. Because three N-shaped fiducials are used, one obtains a system of linear equations that allow for definition of an independent, absolute coordinate system. Therefore one does not have to QA the table indexing of the CT scanner, because the slice position of each scan can be uniquely determined.

Figure 17-4 Phantom with head ring as well as the Brown-Roberts-Wells (BRW) coordinate system in place.

The origin of the BRW system lies within the localizer and therefore divides the brain into eight distinct quadrants. Any point within this geometric configuration can be uniquely defined by specifying anterior-posterior, lateral, and axial coordinates. During the stereotactic CT scan, portions of these N-shaped structures are included in each image, which allows for their identification during treatment planning. To reduce the error in localization, the slice thickness of the CT scan should be the smallest attainable on the scanner. Once the stereotactic CT scan has been acquired, it is transferred to the treatment planning system and the BRW coordinate system is defined by localizing each of the nine rods in each of the CT slices (Fig. 17-5).

Figure 17-5 Four-panel view of acquired stereotactic CT illustrating how the BRW localizer rods are embedded into the CT images.

Next, the target volume, or gross tumor volume (GTV), is outlined. Because magnetic resonance imaging (MRI) often provides better tumor visualization than CT scanning, the GTV is often delineated directly from MRI studies. MRI alone does not provide a sufficient database for stereotactic treatment planning, because magnetic field nonuniformity, gradient field nonlinearity, eddy current effects, and susceptibility artifacts at air/tissue interfaces can introduce significant geometric image distortions that affect the accuracy of treatment plans generated using MRI alone. A three-dimensional volumetric MRI is acquired either before or after head ring placement. After acquiring the stereotactic CT scan, the MRI dataset (which may or may not be stereotactically generated) is fused onto the CT image space through correlation of anatomic landmarks in both image sets. During treatment planning, the regions of interest and dose distribution may be displayed on either the CT image or the fused MRI images, but dosimetric calculations are ideally performed on the underlying CT database (Fig. 17-6). If the MRI is to be obtained after head frame placement, magnetic resonance–compatible head rings need to be used.

Figure 17-6 Image fusion of a stereotactic CT with a T1-3D-SPGR MRI image set with gadolinium contrast. The MRI has been pixilated to show the underlying CT. The difference in tumor visualization between the two imaging modalities is visible.

Some investigators use a “noninvasive” fiducial system in which several fiducials are either implanted subcutaneously or taped on the patient. The fiducials then provide the frame of reference for obtaining stereotactic coordinates. The noninvasive nature of this approach is appealing, but there are concerns regarding perceived problems with patient immobilization during treatment. One such frameless system is the CyberKnife, which is particularly suited for both central nervous system and extracranial applications. The CyberKnife combines a compact 6-MV linear accelerator waveguide with a robotic arm with 6 degrees of freedom. It contains adjustable collimator cones ranging from 5 to 60 mm to deliver highly conformal treatment through several hundred narrow radiation beams from many different angles. For brain radiosurgery, immobilization uses only a thermoplastic facemask. Two orthogonal x-ray cameras mounted to the ceiling allow for real-time tracking based upon bony landmarks or implanted fiducials.

Conventional linac radiosurgery uses multiple non-coplanar arc sets to treat the target volume through a variety of collimator diameters ranging in size from 4 mm to 50 mm, depending on the system (Fig. 17-7). For irregularly shaped lesions such as vestibular schwannomas and arteriovenous malformations (AVMs), arcs delivered through a single circular collimator would lead to the inclusion of a large amount of normal brain tissue, which would yield inferior conformality. In these cases, it is advantageous to use multiple isocenters to conform the prescription isodose shell to the GTV (Fig. 17-8). The dose is typically prescribed to the 70% to 90% isodose line. Whenever multiple isocenters are used, homogeneity is sacrificed for conformality. Ideally, a desirable conformality index (prescribed isodose volume divided by target volume) should be 2 or less and the heterogeneity index (maximum dose divided by prescribed dose) should also be 2 or less.

Figure 17-7 Typical five nonopposing arrangements used in multi-isocenter treatment planning for linac-based radiosurgery.

Figure 17-8 Three selected axial cuts through a vestibular schwannoma showing the resulting conformal treatment plan using four different isocenters. The gadolinium contrast-enhanced thin-slice T1-3D-SPGR MRI data set has been fused to the underlying stereotactic CT to visualize the vestibular schwannoma. The 12.5-Gy treatment isodose line and the 10-Gy isodose line are shown.

Clinical Applications

Both the GK and linac-based forms of radiosurgery have been used in a variety of benign and malignant diseases. Table 17-2 summarizes the most common clinical applications for radiosurgery. In this section we outline some of the basic applications of radiosurgery and results of therapy. Specifics pertaining to each disease are discussed in greater detail in its respective chapter.

Benign Tumors

Meningiomas

Benign diseases make up a considerable proportion of the applications for radiosurgery, with meningioma being one of the most common indications. Up to one-third of meningiomas, however, present in locations not amenable to complete removal. Radiosurgery plays an important role in the treatment of small lesions (<3 to 4 cm) that are surgically inaccessible, such as those in cavernous sinus or posterior parasagittal locations, or those that have been subtotally resected but consistently have been shown to result in high recurrence rates.

Two of the largest series that have examined results of SRS are from the Mayo Clinic and the University of Pittsburgh, both of which have shown local control in more than 90% for benign meningiomas at 5 years.^13,^14,¹⁵ The Mayo Clinic series had a significant complication rate of 13% versus 7% for the University of Pittsburgh series.

Based on published reports, the optimal dose to the margin of the tumor appears to be 12 to 16 Gy. A recent report by Kollova and colleagues ¹⁶ showed that doses above 16 Gy increased the risk of treatment-related edema without a corresponding improvement in tumor control. Inclusion of the dural tail within the treatment volume remains a topic of controversy. Limited histopathologic analysis has shown that microscopic tumor extension into the dural tail is limited to 2 mm.¹⁷ The most common toxicities include cranial nerve deficits for basal tumors and peritumoral edema for nonbasal tumors. The risk of optic neuropathy is very low, with maximum dose constraints to the optic nerves and chiasm of between 8 and 10 Gy.^13,¹⁸ In addition to the dose, a treatment volume of more than 5 cm³, a tumor-brain contact interface of 1 cm² or more, the presence of pretreatment edema, and parasagittal location each increase the risk of peritumoral edema.^16,19,20,21 These factors should be considered when deciding whether or not to include none of the dural tail or only a portion of it within the clinical target volume (CTV), which for all practical purposes is the same as the GTV. With SRS, radiologic response is seen in 50% to 60% of patients, and the remainder of those without progression demonstrated tumor stability.^16,22

Based on these data, it is reasonable to conclude that small meningiomas can be controlled with radiosurgery in the majority of patients, with initial results comparable to those of complete resection (Fig. 17-9). It must, however, be borne in mind that most cases in which this approach has been used have had relatively short follow-up studies and that long-term results (more than 10 years of follow-up studies for each patient) are indeed very sparse. The University of Pittsburgh recently updated their 18-year experience in a cohort of 972 patients. Local control rates in grade 1 meningioma or lesions without histology were 91% and 95%, respectively, in 75 patients with a minimum follow up of 10 years.