Steven D. Chang

THE CYBERKNIFE : A FRAMELESS ROBOTIC SYSTEM FOR RADIOSURGERY

The Cyberknife is a unique instrument for performing frameless stereotactic radiosurgery. Rather than using rigid immobilization, the Cyberknife relies on an image-to-image correlation algorithm for target localization. Furthermore, the system utilizes a novel, light-weight, high-energy radiation source. The authors describe the technical specifications of the Cyberknife and summarize the initial clinical experience.

Image-guided robotic radiosurgery

PURPOSE: To describe the design and performance of a novel frameless system for radiosurgery. This technology, called image-guided radiosurgery (IGR), eliminates the need for stereotactic frame fixation by relating the identified lesion to radiographic landmarks. CONCEPT: IGR uses a lightweight x-band linear accelerator, computer-controlled robotic arm (Fanuc manipulator [Fanuc Robotics North America, Inc., Rochester Hills, MI]), paired orthogonal x-ray imagers, and a computer workstation that performs rapid image-to-image registration. During radiosurgery, the x-ray imaging system determines the location of the lesion and communicates these coordinates to the robot, which adjusts the pointing of the linear accelerator beam to maintain alignment with the target. RATIONALE: Existing stereotactic techniques require rigid cranial fixation to establish and maintain a system of reference for targeting. Such frames cause pain for the patient, limit the use of fractionation, and necessitate a prolonged period of general anesthesia if children are to be treated. Furthermore, skeletal or any other type of rigid fixation is difficult to achieve beyond the cranium. IGR was designed to overcome these limitations, which are inherent to nearly all current radiosurgical methods. DISCUSSION: Preliminary testing and early clinical experience have demonstrated the practicality and potential of the IGR concept and have identified the most important directions for improvement. For example, an IGR prototype accurately tracked target displacements in three dimensions but showed reduced accuracy when confronted by rotational movements. This observation led to development of a new generation of tracking algorithm that promises to improve tracking in all six dimensions. Further experience indicated that improvements in the quality of the x-ray images were needed to allow the system to locate and treat target sites outside the cranium. Consequently, a new x-ray imaging technology with superior resolution and increased sensitivity has been added to the system. These improvements should make it possible to apply IGR techniques to a variety of targets located throughout the body. This article describes and critiques the components of the IGR and summarizes our preliminary clinical experience.

An analysis of the accuracy of the CyberKnife: a robotic frameless stereotactic radiosurgical system.

OBJECTIVEThe use of stereotactic radiosurgical systems to treat intracranial and extracranial tumors and other lesions requires a high degree of accuracy in target identification and localization. The purpose of this study was to evaluate the total system accuracy of the CyberKnife (Accuray, Inc., Sunnyvale, CA), a frameless, image-guided, stereotactic radiosurgery system. METHODSClinically relevant accuracy or application accuracy of the CyberKnife radiosurgery system is based on 1) the beam delivery accuracy, which combines the robot and the camera image tracking system, and 2) target localization accuracy, which combines computed tomographic (CT) imaging and treatment planning. Clinically relevant accuracy can be measured by delivering a radiation dose to phantoms, in which the target is defined on a set of CT images using all components of the CyberKnife system, including the treatment planning software, the robot, the camera tracking system, and the linear accelerator. Clinically relevant accuracy was measured in head phantoms loaded with packs of radiochromic film. The accuracy measured is the displacement of the dose contours from the treatment plan to that measured in the radiosurgically exposed phantom. RESULTSMeasurements of mean errors of the second-generation CyberKnife system at Stanford University Medical Center, installed in 2001, ranged from 0.7 mm for a CT slice thickness of 0.625 mm to 1.97 mm for a CT slice thickness of 3.75 mm. CONCLUSIONThe frameless, image-guided, second-generation CyberKnife radiosurgery system has a clinically relevant accuracy of 1.1 ± 0.3 mm when CT slice thicknesses of 1.25 mm are used. CyberKnife precision is comparable to published localization errors in current frame-based radiosurgical systems.

Image-guided Hypo-fractionated Stereotactic Radiosurgery to Spinal Lesions

OBJECTIVEThis article demonstrates the technical feasibility of noninvasive treatment of unresectable spinal vascular malformations and primary and metastatic spinal tumors by use of image-guided frameless stereotactic radiosurgery. METHODSStereotactic radiosurgery delivers a high dose of radiation to a tumor volume or vascular malformation in a limited number of fractions and minimizes the dose to adjacent normal structures. Frameless image-guided radiosurgery was developed by coupling an orthogonal pair of x-ray cameras to a dynamically manipulated robot-mounted linear accelerator that guides the therapy beam to treatment sites within the spine or spinal cord, in an outpatient setting, and without the use of frame-based fixation. The system relies on skeletal landmarks or implanted fiducial markers to locate treatment targets. Sixteen patients with spinal lesions (hemangioblastomas, vascular malformations, metastatic carcinomas, schwannomas, a meningioma, and a chordoma) were treated with total treatment doses of 1100 to 2500 cGy in one to five fractions by use of image-guided frameless radiosurgery with the CyberKnife system (Accuray, Inc., Sunnyvale, CA). Thirteen radiosurgery plans were analyzed for compliance with conventional radiation therapy. RESULTSTests demonstrated alignment of the treatment dose with the target volume within ± 1 mm by use of spine fiducials and the CyberKnife treatment planning system. Tumor patients with at least 6 months of follow-up have demonstrated no progression of disease. Radiographic follow-up is pending for the remaining patients. To date, no patients have experienced complications as a result of the procedure. CONCLUSIONThis experience demonstrates the feasibility of image-guided robotic radiosurgery for previously untreatable spinal lesions.

Staged stereotactic irradiation for acoustic neuroma.

OBJECTIVE: Stereotactic radiosurgery has proven effective in the treatment of acoustic neuromas. Prior reports using single-stage radiosurgery consistently have shown excellent tumor control, but only up to a 50 to 73% likelihood of maintaining hearing at pretreatment levels. Staged, frame-based radiosurgery using 12-hour interfraction intervals previously has been shown by our group to achieve excellent tumor control while increasing the rate of hearing preservation at 2 years to 77%. The arrival of CyberKnife (Accuray, Inc., Sunnyvale, CA) image-guided radiosurgery now makes it more practical to treat acoustic neuroma with a staged approach. We hypothesize that such factors may further minimize injury of adjacent cranial nerves. In this retrospective study, we report our experience with staged radiosurgery for managing acoustic neuromas. METHODS: Since 1999, the CyberKnife has been used to treat more than 270 patients with acoustic neuroma at Stanford University. Sixty-one of these patients have now been followed up for a minimum of 36 months and form the basis for the present clinical investigation. Among the treated patients, the mean transverse tumor diameter was 18.5 mm, whereas the total marginal dose was either 18 or 21 Gy using three 6- or 7-Gy fractions. Audiograms and magnetic resonance imaging were obtained at 6-months intervals after treatment for the first 2 years and then annually thereafter. RESULTS: Of the 61 patients with a minimum of 36 months of follow-up (mean, 48 mo), 74% of patients with serviceable hearing (Gardner-Robinson Class 1–2) maintained serviceable hearing at the last follow-up, and no patient with at least some hearing before treatment lost all hearing on the treated side. Only one treated tumor (2%) progressed after radiosurgery; 29 (48%) of 61 decreased in size and 31 (50%) of the 61 tumors were stable. In no patients did new trigeminal dysfunction develop, nor did any patient experience permanent injury to their facial nerve; two patients experienced transient facial twitching that resolved in 3 to 5 months. CONCLUSION: Although still preliminary, these results indicate that improved tumor dose homogeneity and a staged treatment regimen may improve hearing preservation in acoustic neuroma patients undergoing stereotactic radiosurgery.

Multimodality treatment of giant intracranial arteriovenous malformations

OBJECTIVE Giant arteriovenous malformations (AVMs) (i.e., those greater than 6 cm at maximum diameter) are difficult to treat and often carry higher treatment morbidity and mortality rates than do smaller AVMs. In this study, we reviewed the treatment, angiographic results, and clinical outcomes in 53 patients with giant AVMs who were treated at Stanford between 1987 and 2001. METHODS The patients selected included 20 males (38%) and 33 females (62%). Their presenting symptoms were hemorrhage (n = 20; 38%), seizures (n = 18; 34%), headaches (n = 8; 15%), and progressive neurological deficits (n = 7; 13%). One patient was in Spetzler-Martin Grade III, 9 were in Spetzler-Martin Grade IV, and 43 were in Spetzler-Martin Grade V. The mean AVM size was 6.8 cm (range, 6-15 cm). AVM venous drainage was superficial (n = 7), deep (n = 20), or both (n = 26). At presentation, 31 patients (58%) were graded in excellent neurological condition, 17 were graded good (32%), and 5 were graded poor (9%). RESULTS The patients were treated with surgery (n = 27; 51%), embolization (n = 52; 98%), and/or radiosurgery (n = 47; 89%). Most patients received multimodality treatment with embolization followed by surgery (n = 5), embolization followed by radiosurgery (n = 23), or embolization, radiosurgery, and surgery (n = 23). Nineteen patients (36%) were completely cured of their giant AVMs, 90% obliteration was achieved in 4 patients (8%), less than 90% obliteration was achieved in 29 patients (55%) who had residual AVMs even after multimodality therapy, and 1 patient was lost to follow-up. Of the 33 patients who either completed treatment or were alive more than 3 years after undergoing their most recent radiosurgery, 19 patients (58%) were cured of their AVMs. The long-term treatment-related morbidity rate was 15%. The clinical results after mean follow-up of 37 months were 27 excellent (51%), 15 good (28%), 3 poor (6%), and 8 dead (15%). CONCLUSION The results in this series of patients with giant AVMs, which represents the largest series reported to date, suggest that selected symptomatic patients with giant AVMs can be treated successfully with good outcomes and acceptable risk. Multimodality treatment is usually necessary to achieve AVM obliteration.

Microsurgical resection of brainstem, thalamic, and basal ganglia angiographically occult vascular malformations.

OBJECTIVE To evaluate the clinical results for patients who underwent resection of angiographically occult vascular malformations (AOVMs) of the brainstem, thalamus, or basal ganglia, successfully resected after it exhibited rebleeding and presented to a pial surface. METHODS Between January 1990 and May 1998, 56 patients with 57 deep AOVMs underwent 63 operations, at Stanford University Medical Center, to treat AOVMs of the brainstem (42 AOVMs), thalamus (5 AOVMs), or basal ganglia (10 AOVMs). The surgical approach was suboccipital midline (27 operations), far lateral suboccipital (10 operations), transsylvian (9 operations), interhemispheric transcallosal or infracallosal (8 operations), infratentorial supracerebellar (6 operations), or subtemporal (3 operations). Four patients experienced recurrent bleeding from the same lesion after surgical resection, requiring a second operation. One patient required a planned second operation, using a different approach, to completely resect the lesion, and one patient underwent two surgical procedures to resect two separate brainstem AOVMs. One patient initially underwent exploration but not resection of her AOVM, because it did not present to a pial or ependymal surface. The AOVM was successfully resected after it exhibited rebleeding and presented to a pial surface. RESULTS The immediate outcomes after surgery were unchanged for 31 patients (55%), worsened for 16 (29%), and improved for 9 (16%). The long-term outcomes were unchanged for 24 patients (43%), compared with their presenting grade, worse for 3 (5%), and improved for 29 (52%). Patients who had undergone previous radiotherapy or radiosurgery to treat these lesions experienced more difficult postoperative courses, and radiation necrosis was observed for two patients. CONCLUSION AOVMs of the brainstem, thalamus, and basal ganglia can be safely removed, with a long-term neurological morbidity rate of only 5% and a complete lesion resection rate of 93% after the initial planned resection. The use of cranial base surgical approaches and intraoperative electrophysiological monitoring contributes to successful clinical outcomes.

The cerebrospinal fluid production rate is reduced in dementia of the Alzheimer’s type

Objective: To evaluate the production rate of CSF in patients with differing disease states. Methods: The authors measured the production rate of CSF in three groups of patients: five patients with PD below age 60 (aged 51 ± 4 years, mean ± SD), nine with PD over age 60 (aged 69 ± 6 years, mean ± SD), and seven with dementia of the Alzheimer’s type (AD) (aged 72 ± 9 years, mean ± SD). This method, based on the Masserman technique, employs ventricular rather than a lumbar access to the CSF space. Furthermore, the volume of CSF removed during the procedure is only 3 mL rather than 10 mL. Results: These measurements indicate that the mean rate of CSF production in patients with PD under age 60 was 0.47 ± 0.13 mL/minute, in patients with PD aged 60 or older the mean rate was 0.40 ± 0.12 mL/minute, and in patients with AD the mean rate was 0.20 ± 0.06 mL/minute. Conclusion: These results indicate that the rate of CSF production in patients with PD is normal, and that the rate of CSF production in patients with AD is markedly reduced.

Visual field preservation after multisession cyberknife radiosurgery for perioptic lesions.

OBJECTIVE:The restricted radiation tolerance of the anterior visual pathways represents a unique challenge for ablating adjacent lesions with single-session radiosurgery. Although preliminary studies have recently demonstrated that multisession radiosurgery for selected perioptic tumors is both safe and effective, the number of patients in these clinical series was modest and the length of follow-up limited. The current retrospective study is intended to help address these shortcomings. METHODS:Forty-nine consecutive patients with meningioma (n = 27), pituitary adenoma (n = 19), craniopharyngioma (n = 2), or mixed germ cell tumor (n = 1) situated within 2 mm of a “short segment” of the optic apparatus underwent multisession image-guided radiosurgery at Stanford University Medical Center. Thirty-nine of these patients had previous subtotal surgical resection, and six had previously been treated with conventional fractionated radiotherapy (6). CyberKnife radiosurgery was delivered in two to five sessions to an average tumor volume of 7.7 cm3 and a cumulative average marginal dose of 20.3 Gy. Formal visual testing and clinical examinations were performed before treatment and at follow-up intervals beginning at 6 months. RESULTS:After a mean visual field follow-up of 49 months (range, 6–96 mo), vision was unchanged postradiosurgery in 38 patients, improved in eight (16%), and worse in three (6%). In each instance, visual deterioration was accompanied by tumor progression that ultimately resulted in patient death. However, one of these patients, who had a multiply recurrent adrenocorticotropic hormone-secreting pituitary adenoma, initially experienced early visual loss without significant tumor progression after both a previous course of radiotherapy and three separate sessions of radiosurgery. After a mean magnetic resonance imaging follow-up period of 46 months, tumor volume was stable or smaller in all other cases. Two patients died of unrelated nonbrain causes. CONCLUSION:Multisession radiosurgery resulted in high rates of tumor control and preservation of visual function in this group of perioptic tumors. Ninety-four percent of patients retained or improved preradiosurgical vision. This intermediate-term experience reinforces the findings from earlier studies that suggested that multisession radiosurgery can be a safe and effective alternative to either surgery or fractionated radiotherapy for selected lesions immediately adjacent to short segments of the optic apparatus.

Patterns of patient movement during frameless image-guided radiosurgery

PURPOSE Image-guided radiosurgery aligns the treatment beam to the target site by using a radiographic imaging system to locate anatomic landmarks associated with the treatment target. Because the procedure is performed without a rigid frame, the precision of dose alignment can be affected by patient movement. Movement is limited by noninvasive restraints and compensated by remeasuring the target position at short intervals throughout treatment and then realigning the beam. Frameless image-guided radiosurgery has been used at our institution to treat 250 cranial, 23 spinal, 9 lung, and 3 pancreas cases involving malignant and benign tumors as well as vascular malformations. We have analyzed the target position records for all of these cases to assess the frequency, magnitude, and case-by-case patterns of patient movement. METHODS AND MATERIALS The position of the treatment site during image-guided radiosurgery was measured at approximately 1-2-min intervals, on average, using orthogonal amorphous silicon X-ray cameras and an image registration process that determined all six degrees of freedom in the targets position. The change in position from one measurement to the next was indicative of patient movement. RESULTS The treatment site position along each axis of translation was observed to vary by an average of 0.45 mm for the cranium, 0.53 mm for the cervical spine, 0.53 mm for the lumbar and thoracic spine, 1.06 mm for the lung, and 1.50 mm for the pancreas. Half of all cranial cases showed systematic drifting of the target away from the initial setup position. CONCLUSION Using noninvasive restraints and supports, short-term movement of the head and spine during image-guided radiosurgery was limited to a radius of 0.8 mm, which satisfies the prevailing standard for radiosurgical dose alignment precision, but maintaining this margin of error throughout a treatment fraction requires regular monitoring of the target sites position.