Stereotactic radiosurgery has transformed the practice of neurosurgery and, over the past decade, has increasingly affected all surgical disciplines. First-generation radiosurgical devices used stereotactic frames for both localizing targets within the brain and immobilizing the head during treatment. This basic design limited radiosurgery to the treatment of brain lesions where the rigidity of the skull provided adequate skeletal purchase. To transcend such limitations, image-guided radiosurgery was developed. In lieu of the frame, near–real-time image guidance by means of precisely calibrated diagnostic x-rays provides a methodology for head and, by extrapolation through computerized tomography, target localization. Although image-guided targeting is very flexible, working virtually anywhere in the body, this frameless approach does not provide a direct mechanism for patient immobilization; image guidance can detect patient movement but by itself does not permit compensation of such offsets. Although one could envision an approach to image-guided radiosurgery that required frequent manual repositioning of the patient to accommodate inevitable “involuntary” patient movements, such a method would seem a priori to be rather slow and lack robustness. In the setting of radiosurgery, robotic retargeting of the therapeutic x-ray source proved to be the optimal technical accompaniment to image guidance. This combination of image guidance and robotics enables dynamic compensation for patient movement, which in turn has ushered in a transformation of radiosurgical technology and clinical applications.
The above real-time feedback loops ensured both theoretically and in practice that radiosurgery remained extremely precise. This design provided the technical foundation for the first robotic radiosurgical device, the CyberKnife (Accuray Inc, Sunnyvale, California; Figure 1), and much of the field of image-guided radiation that subsequently evolved.1 There are no universally accepted measures of radiosurgical accuracy, but generally, total system errors of < 1 mm are deemed acceptable for nearly all clinical applications. Over the past decade, this degree of clinical fidelity has been repeatedly confirmed for image-guided robotic radiosurgery, whether using an “off-the-shelf” industrial robot as does the CyberKnife system or for a range of more customized and more general-purpose therapeutic radiation machines. After more than a decade of experience with robotic radiosurgery, outcomes for all clinical indications are now widely accepted to be comparable to those of frame-based procedures.2
The unique capabilities of image guidance allowed the emergence of a number of new clinical applications in the brain. Multisession radiosurgery for perioptic lesions and large tumor resection cavities is a case in point. Equally important, robotic radiosurgery has opened up new vistas in treating spinal lesions, a field of treatment that is now evolving quickly. Nevertheless, the most dramatic impact of image-guided robotic radiosurgery has been the emergence of a vast array of new nonneurosurgical clinical applications within the chest, abdomen, and pelvis. In fact, today no anatomic location is outside the scope of practice for modern robotic radiosurgery techniques.
CLINICAL APPLICATIONS IN NEUROSURGERY
Given the tremendous success of robotic radiosurgery over the past decade, where might technology be headed now? The equipment of the future should reflect the demands of the evolving clinical applications. What are those indications? For starters, there is growing interest and evidence to support the treatment of large numbers of brain metastases rather than the limited number (< 3) of lesions that characterized clinical practice a decade ago. As a consequence, the time required for treating ≥ 10 metastases, for example, can start to pose practical constraints on the physician’s decision to use or not use stereotactic radiosurgery in some patients. Therefore, it seems likely that robotic radiosurgical instruments will, in the future, need to be able to accommodate such opportunities. In some ways, this future has started to arrive already. The Varian Medical Systems (Palo Alto, California) Rapid Arc enables large numbers of brain metastase to be treated in a single rotational arc around a patient’s head with a high measure of conformality (Figure 2A-C). Such treatment is enabled by a very-high-output linear accelerator (the beam-flattening filter has been removed) with an accompanying large radiation field, an elaborate computer-controlled multileaf collimator used to shape the beam, and very fast computerized arc-based volumetric treatment planning strategies. Multileaf collimators with even thinner leaves and driven by even faster motors, as well as both better dose optimization algorithms and the use of noncoplanar arcs, will in the future enable even larger numbers of brain lesions to be treated with still less dose to normal brain.
The use of robotic radiosurgery to treat spinal tumors is likely to grow considerably in the near future. Greater experience and more clinical outcome studies are driving such opportunities. One of the shortcomings of the CyberKnife robotic system, the first-generation spinal radiosurgery technology, was the limited ability to direct beams at a target from posteriorly. As a result, entering beams cause radiation-sensitive abdominal structures (ie, small bowel) to be disproportionately irradiated. Although prone CyberKnife treatments are possible, arc-based treatments with extremely fast image-guided conventional linear accelerators or recently available dedicated devices like the Vero (BrainLAB GmbH, Munich, Germany) can obviate much of this concern by more readily enabling posterior treatment beams. When combined with better multileaf collimators and treatment planning software, the efficacy and safety of image-guided robotic radiosurgery will displace most spinal radiotherapy and in many cases the need for more open surgical resection of spinal lesions.
Dedicated radiosurgery systems, beginning with the Gamma Knife (Elekta, Stockholm, Sweden), have always proven to be particularly attractive to surgeons. Experience with the Gamma Knife has shown that by focusing on a more narrow set of clinical applications, instrument design can be optimized and provide maximal precision and conformality. However, more limited flexibility tends to add appreciable cost to a specialized device, with the Gamma Knife and the Vero system being cases in point. Could it be possible to make specialized robotic radiosurgical devices that are low cost? Current robotic devices have proven to be relatively complex, requiring considerable technical support. Might it also be possible to engineer simplicity into complex radiosurgical instruments, thereby making them intuitive even for neurosurgeons not subspecialty trained? Given the relatively modest demand that stems from existing robotic radiosurgical clinical indications, it is hard, from an economic standpoint, to justify the sizable investment required to create both more intuitive and cheaper specialized devices for neurosurgeons. However, might there be new higher-volume clinical applications on the horizon? It is a credible speculation that behavioral and psychiatric disorders will be a new class of disease for which precision radiation could have substantial benefits. Until now, radiosurgery has been used exclusively and sparingly as a noninvasive method for ablating brain circuits underlying certain psychiatric diseases, making it merely a more palatable successor to frontal lobotomy. Such an approach to treating these almost ubiquitous disorders seems primitive next to the burgeoning field of neuromodulation that has grown up in recent years around deep brain stimulation. However, the concept of “radiomodulation,” which leverages the peculiar radiobiology of irradiated brain circuits, could represent an important noninvasive alternative to “invasive” deep brain surgery.3,4 If so, such functional procedures could drive future technical innovation within robotic radiosurgery for brain applications.
NONNEUROSURGICAL ROBOTIC RADIOSURGERY
Despite the continued evolution of radiosurgery within the brain and spine, the biggest opportunities for robotic radiosurgery continue to lie within nonneurosurgical specialties. Because real-time sensing of target location inside the chest and abdomen was not possible 2 decades ago when robotic radiosurgery was created (ie, the CyberKnife), the best targeting method of the time was used: periodic orthogonal kilovoltage imaging of embedded x-ray opaque markers. With this design, the position of the tumor is inferred with respect to the markers. Between episodic x-rays, target position was surmised by computer models formulated from recently acquired historic data. Because lesions within the chest and abdomen move significantly from respiratory and cardiac motions, automated real-time robotic compensation for these types of movements became a key enabling technology.
There is an element of cyclicality to respiratory and cardiac movements that by virtue of being predictable can be compensated for by automated means. Present-day robotic radiosurgery compensates for respiratory motion in real time by either dynamically retargeting the linear accelerator beam or turning the treatment beam on only when the target is at a preestablished position such as end expiratory. However, there remains an element of randomness and unpredictability of some movements related to the respiratory cycle. Given the nature of respiration, the frequency of breathing and the depth of inspiration/expiration are constantly varying. Moreover, varying tissue viscoelasticity, regional variations within organs, and the nature of lung hysteresis make modeling and predicting the location of a tumor within the chest and abdominal an ongoing challenge.
Better software algorithms for modeling soft tissue displacement throughout respiration could improve the accuracy of localizing a radiosurgical lesion within the chest or abdomen. However, dynamic sensing of tumor motion, if combined with robotic feedback, could obviate this shortcoming, as embodied by a technology that uses tiny electromagnetic transponders.5 With this approach, an external electromagnetic array is used to excite a percutaneously or bronchoscopically embedded transponder, ie a miniature antenna, in a manner that emits radiofrequency waves. Placing one of these transponders near a lesion that is affected by diaphragmatic movement allows the position of the radiosurgical target to be surmised in real time. As long as there is not significant movement of the interposed tissues, the location of the tumor can be surmised with great accuracy both in absolute terms and over time. When linked with a robotic delivery system, real-time compensation for movement of the tumor becomes quite straightforward.
Future sensors, perhaps using other forms of energy like magnetic, ultrasonic, or near infrared, will enable comparable dynamic retargeting of the therapeutic beam. In this regard, new radiation delivery systems that use magnetic resonance imaging to localize moving tumors like lung cancer (ViewRay, Inc, Cleveland, Ohio) are just starting to arrive. Regardless, robotic effectors, defined broadly, are key to making all such concepts work. The Holy Grail remains the ability to follow and treat robotically even fast-moving structures like the heart with external radiation. Why target the heart? Compelling animal studies suggest that cardiac ablation with radiosurgery could substitute for more invasive and less permanent radiofrequency ablation of the posterior atrium as a treatment for atrial fibrillation.6 Because cardiac arrhythmias are the biggest cause of stroke in many regions of the world, the success of such a technology could have profound implications for the clinical practice of all neurosurgeons. As has been true throughout the history of radiosurgery, the arrival of technology is only the first step in understanding the future clinical applications.
It is important to note that a robot performing radiosurgery need not be an articulated arm. Radiosurgical robots should be broadly defined as any autonomous or semiautonomous electromechanical machine that is guided by computer and electronic programming in the performance of the task of tissue ablation. The emergence of newer therapeutic x-ray sources in the future may call for even newer concepts of robots. Microbeam irradiation, generated by extremely expensive synchrotron light sources, has some astonishing biologic properties that permit high ablation efficacy while preserving normal tissues.7 If more compact and affordable microbeam sources can be developed in the future, one could readily envision the need for complementary robotic targeting systems. Another futuristic x-ray source may involve laser generation of intense electron or proton beams inside patients using tiny (2 mm) accelerators. If such technology can be developed, a totally new generation of robots will be required to precisely advance both the accelerator and an attached fiberoptic catheter that serves to transmit an intense laser beam.8 The concept of robotic radiosurgery is likely to take many shapes going forward. Needless to say, efforts to predict the future of anything are fraught with fatuity. Therefore, when all is said and done, the best way to predict the future (of robotic radiosurgery) is, to paraphrase Alan Kay, to invent it.
At the heart of all of surgery lies one or both fundamental objectives: resection (lesion removal, abscess or clot drainage, etc) or reconstruction. The future of reconstructive surgery is unclear, but it seems highly possible that most resection, outside acute situations like trauma, will be managed by means of noninvasive target-directed energy. As evidenced by the widespread use of radiosurgery, this change is already underway. However, further progress will hinge on perceptions among surgeons: Who are they, and how do they carry out their mission? Throughout history, an operation consisted of extemporaneous manual processes during which surgeons would navigate a patient’s internal anatomy. Modern tools for computerized navigation have led to some recent reinterpretation of the nature of surgery, but most operations still remain unrehearsed free-form events in which critical judgments are made in real time. Going forward, this balance of skills will change. The tools of noninvasive surgery, as manifested by radiosurgery, allow far less manual manipulation. However, to enable the next stage of the evolution of surgery, robotic systems will assume even greater control of all mechanical processes, thereby automating the entire process of radiosurgery. Although surgeons will remain critical to the process of comprehensive pretreatment planning, once properly programmed, robots will be the end effectors within radiosurgery. In doing so, robots will constitute a metaphorical extension of the surgeon’s hands and thereby remind future generations of the manual origins of surgery.
Dr Adler is an employee of Stanford University and an employee and shareholder of Varian Medical Systems, Inc.
1. Adler JR, Chang SD, Murphy MJ, Doty J, Geis P, Hancock SL. The CyberKnife: a frameless robotic system for radiosurgery. Stereotactic Funct Neurosurg. 1998;69(1-4 pt 2):124–128.
2. Ho A, Fu D, Cotrutz C, et al.. A study of the accuracy of CyberKnife spinal radiosurgery using skeletal structure tracking. Neurosurgery. 2007;60(2 suppl 1):ONS147–ONS156.
3. Pellmar TC, Schauer DA, Zeman GH. Time- and dose-dependent changes in neuronal activity by x-radiation in brain slices. Radiat Res. 1990;122(2), 209–214.
4. Schneider B, Borchers JD, Adler JR. Radiation-based neuromodulation: rationale and new directions. Cureus. 2010;2(2):e8.
5. Wu J, Ruan D, Cho B, et al.. Electromagnetic detection and real-time DMLC adaptation to target rotation during radiotherapy. Int J Radiat Oncol Biol Phys. 2012;82(3):e545–e553.
6. Maguire PJ, Gardner E, Jack AB, et al.. Cardiac radiosurgery (CyberHeart™) for treatment of arrhythmia: physiologic and histopathologic correlation in the porcine model. Cureus. 2011;3(8):e32.
7. Romanelli P, Fardone E, Brauer-Krisch E, Prezado Y, Bravin A. Emerging neurosurgical applications of synchrotron-generated microbeams. Cureus. 2011;3(7):e29.
8. Esarey E, Schroeder BC, Leemans WP. Physics of laser-driven plasma-based electron accelerators. Rev Mod Phys. 2009;81(3):1269–1280.