Quality Assurance Results for a Commercial Radiosurgery System: A Communication

Technology in Cancer Research and Treatment ISSN 1533-0346 Volume 14 Number 5 October 2015 2014 June 16. Epub ahead of print. Quality Assurance Results for a Commercial Radiosurgery System: A Communication www.tcrt.org DOI: 10.7785/tcrt.2012.500433 The purpose of this communication is to inform the radiosurgery community of quality assurance (QA) results requiring attention in a commercial FDA-approved linac-based cone stereotactic radiosurgery (SRS) system. Standard published QA guidelines as per the American Association of Physics in Medicine (AAPM) were followed during the SRS system s commissioning process including end-to-end testing, cone concentricity testing, image transfer verification, and documentation. Several software and hardware deficiencies that were deemed risky were uncovered during the process and QA processes were put in place to mitigate these risks during clinical practice. In particular, the present work focuses on daily cone concentricity testing and commissioning-related findings associated with the software. Cone concentricity/alignment is measured daily using both optical light field inspection, as well as quantitative radiation field tests with the electronic portal imager. In 10 out of 36 clini- cal treatments, adjustments to the cone position had to be made to align the cone with the collimator axis to less than 0.5 mm and on two occasions the pre-adjustment measured offset was 1.0 mm. Software-related errors discovered during commissioning included incorrect transfer of the isocentre in DICOM coordinates, improper handling of non-axial image sets, and complex handling of beam data, especially for multi-target treatments. QA processes were established to mitigate the occurrence of the software errors. With proper QA processes, the reported SRS system complies with tolerances set out in established guidelines. Discussions with the vendor are ongoing to address some of the hardware issues related to cone alignment. Mark Ruschin, Ph.D. 1,2 * Alexander Lightstone, Ph.D. 1,2 David Beachey, Ph.D. 1,2 Matt Wronski, Ph.D. 1 Steven Babic, Ph.D. 1 Collins Yeboah, Ph.D. 1,2 Young Lee, Ph.D. 1 Hany Soliman, M.D. 2,3 Arjun Sahgal, M.D. 2,3 1 Department of Medical Physics, Sunnybrook Odette Cancer Centre, Toronto, ON, Canada 2 Department of Radiation Oncology, University of Toronto, Toronto, ON, Canada 3 Department of Radiation Oncology, Sunnybrook Odette Cancer Centre, Toronto, ON, Canada Key words: Image-guidance; Radiosurgery; Stereotactic; Stereotactic cones. Introduction The AAPM Report 54 (Task Group 42) provides a detailed description of stereotactic radiosurgery (SRS) systems, sources of error, and quality assurance (QA) tolerances (1). For example, according to Report 54, all machine axes should coincide within a 1 mm radius sphere for all possible gantry, collimator and couch angles (1). More recent practice guidelines such as AAPM TG-142 and ACR guidelines have updated some tests, such as those related to cone-beam CT image guidance and frameless radiosurgery (2). Additionally, Canadian guidelines from the Canadian Abbreviations: AAPM: American Associate of Physicists in Medicine; ACR: American College of Radiology; CAPCA: Canadian Association of Provincial Cancer Agencies; CT: Computed Tomography; DICOM: Digital Imaging and Communications in Medicine; EPID: Electronic Portal Imaging Device; FDA: Food and Drug Administration; MR: Magnetic Resonance; MU: Monitor Unit; QA: Quality Assurance; R&V: Record and Verify; SRS: Stereotactic Radiosurgery; TPS: Treatment Planning System. *Corresponding author: Mark Ruschin, Ph.D. Phone: 416-480-6400 X 87562 E-mail: Mark.Ruschin@sunnybrook.ca 601

602 Ruschin et al. Association of Provincial Cancer Agencies (CAPCA) have been developed with more specific tolerances and action levels for the various areas within radiation oncology. For example, based on CAPCA report on radiosurgery/radiotherapy, the tolerance for cone-alignment is 0.5 mm (3). In June of 2013 a new SRS system was installed at our institution and consisted of a set of stereotactic cones and an associated treatment planning system (TPS) from a major linac vendor (Elekta AB, Crawley, England). FDA clearance for the TPS system (ERGO++) and stereotactic cones was received in 2008 and 2001 respectively. Between June 2013 and October 2013, we performed a series of standard commissioning tests based on published guidelines, including independent dose verification by the Radiological Physics Center in Houston, Texas. From October 2013 to February 2014, we have used the system to treat 36 patients, comprising 82 brain metastases. The purpose of the present report is to inform the SRS community to look for areas of potential risks that were identified during the commissioning process and provide QA processes that can be utilized to mitigate those risks. Materials and Methods Stereotactic Cone Tray As shown in Figure 1, the cones are screwed into an aluminum tray that is mounted onto the linac head. The tray is on a translational stage that can be adjusted with two perpendicular micrometers. The cone tray is attached to the linac head with a set of four thumb screws, and held in place by friction. During commissioning, the stage was adjusted to centre the cone aperture on the mechanical isocentre under collimator rotation. A daily QA test was then implemented in which the Figure 1: Photograph of the adjustable cone tray. The cone tray is attached to the linac head with four thumb screws (one near each corner) and the central tray containing the cone can be translated using the micrometers (indicated with white arrows). electronic portal imaging device (EPID) captures two images of the circular cone separated by a collimator rotation of 180 degrees, as shown in Figure 2. The measured displacement of the cone aperture between the two images indicates the precession of the cone under collimator rotation. The alignment error between the collimator axis and the cone axis is approximately half the measured distance. Since the flat panel detector remains stationary between the two measurements, the difference between the two images is purely due to collimator precession, resulting from the cone long axis not being aligned with the collimator axis of rotation. The resolution of the images is 0.25 mm projected at the isocentre plane. A measured displacement of greater than 0.4 mm Figure 2: Day-of-treatment cone concentricity test. Left: EPID image of cone at Collimator 5 270. Right: EPID image of cone at Collimator 5 90. The distance from the digital cross-hair in purple (not indicative of the mechanical isocentre) to the edge of the collimator radiation field is measured on both images and the difference between the two measurements should be 0.4 mm, otherwise the cone tray may be adjusted. In this case, the distance is 0.0 mm.

Errors in Commercial Radiosurgery System 603 (i.e., 2 pixels) indicates a cone misalignment of greater than 0.2 mm and requires an investigation and potential adjustment in order to ensure that we are safely within the tolerance recommended in the CAPCA guidelines (3). We report the measured displacement for all 36 treatment days and the frequency of adjustments. Treatment Planning System The treatment planning system (TPS) consists of four modules: (1) Image management, (2) Localization, (3) Image fusion, and (4) Radiosurgery planning. Image datasets: During the commissioning process, we encountered several issues (i.e., software bugs) in terms of how the TPS handles image sets in the image management, image fusion, and radiosurgery planning modules. We report these bugs and how we have mitigated their effect. Transfer of beam data: Once a treatment plan has been generated in the radiosurgery planning module, a TPS report can be generated and the beam data is exported to the record and verify (R&V) system. Several errors/bugs and complications were encountered during the commissioning process that we will report and how we have attempted to mitigate their effects. It should be noted these types of errors tended to be exacerbated when planning for multiple lesions for which the TPS is limited in terms of its ability to handle. Isocentre coordinate for cone-beam CT image guidance: Along with the beam data, the isocentre coordinates in DICOM space are also sent to the R&V system for use with cone-beam CT image guidance. A software bug was encountered during commissioning, which bears reporting as a separate issue apart from the transfer of beam data. Results Stereotactic Cone Tray The displacement of the cone between collimator 5 90 and collimator 5 270 as a function of treatment day is shown in Figure 3. On 10 of 36 clinical days, an adjustment to the cone tray was made using the micrometers to bring the test within a 0.4 mm tolerance. The physical distance to adjust the tray is roughly half of the measured displacement, i.e., if the displacement was 0.5 mm at the isocentre, a shift of approximately 0.25 mm was applied to correct it. The standard deviation in the pre-adjustment displacement value was 0.25 mm and 0.1 mm in the gun-target and left-right positions, respectively. While the measurement of cone displacement with collimator displacement is part of the standard patient QA test, we suggest that there should be mechanical pins to assist in re-locating this cone tray on the linac Figure 3: Plot of cone collimator precession each clinical day. G-T is Gun-Target direction and L-R is left-right. The dashed arrows point to the final value after an adjustment was made to the cone tray to bring the test within tolerance. Of 36 clinical days, adjustments were made on 10 days. head with accuracy better than 0.1 mm. We also recommend a plate to better protect the stage micrometers from accidental manipulation. Treatment Planning System Image datasets: A software bug was encountered in which the TPS accepts images that are tilted with respect to the true axial plane, but does not properly link contours or fusion transformations to this orientation. As a result, if a contour is drawn on a tilted MR image set the contour will appear shifted relative to its initial position the next time the image set is opened, as shown in Figure 4. Similarly, if the tilted MR image set is properly fused to the CT dataset in the Fusion module, the MR image set will appear shifted relative to the CT set the next time the MR is loaded. To mitigate this problem, we have taken steps to ensure that the primary images used for planning are not tilted. The Orientation tag in the DICOM header (0020, 0037) should be 1\0\0\0\1\0 for true axial images. If secondary images (such as MR) are tilted but the primary images (e.g., CT) are not, then make sure to assign all contours to the primary, non-tilted dataset and verify that they are in the correct location within the Fusion module. Transfer of beam data: The treatment plan report generated by the TPS includes (for each arc) the average depth, MU, and the dose at isocentre. Several issues have complicated the process for performing QA of the beam data transferred to the R&V system. First of all, the TPS report truncates the values of MUs and dose values, cutting off anything after the decimal place. As a result, all of MUs loaded up in the R&V differ from the TPS report by up to 0.9 MU per arc. Additional issues arise when multiple targets/lesions are planned. At our institution, planning for two or more brain metastases is common.

604 Ruschin et al. Figure 4: Example of re-opening a tilted image set on contour position. Image on the left is the original contour. The next two images are subsequent saves and re-opening of the images with continual displacement of the contour. Furthermore, each lesion may involve multiple isocentres. However, as the number of isocentres being used increases, it becomes increasingly challenging to safely track these data. The current TPS version assigns only one dose prescription per plan, linked to one isocentre. If a multiple isocentre plan is generated and exported, there is no identification given in the R&V as to which beam corresponds to which isocentre, or what prescription is assigned to any target. To mitigate the risk of inaccurate data transfer, all of our multi-isocentre plans are manually split up into individual plans, i.e., one plan for each isocentre. The composite plan is copied once for each iso, and in each copy all but one isocentre (the treated one) is deleted. The prescriptions are re-normalized to that isocentre. Since the dose per arc is also truncated, when the doses are summed up for a given target, the truncation error per arc also gets summed. Re-prescribing in the individual plans based on the erroneous summed dose has lead to discrepancies in MU of up to 5 MU per arc. Therefore, manual inspection and multiple checks by planning dosimetrist, physicist, and radiation therapist(s) are required. We have therefore developed an extensive and explicit protocol for handling the beam transfer, which involves independent checks that all of the plan data has been properly transferred to the R&V. The independent checks include: MU, jaw-size, couch and gantry angles, site setup (DICOM coordinate transfer), visual verification of isocentre position and distribution on CT/MR images, plan revision number verification, and setup-plates cross-verification with stereotactic coordinates. We also have implemented a secondary MU calculation for each isocentre using measured tissue-phantom-ratio and relative output factor look-up tables, with a tolerance of 2% for the agreement between planning and calculated MUs. Isocentre coordinate for cone-beam CT image guidance: As part of the published guidelines for quality assurance of cone-beam CT it is necessary to verify the transfer of information from the TPS to the treatment unit (4). Since cone-beam CT uses a different coordinate system than stereotactic, verification of the isocentre coordinates is vital. At our institution, we discovered that the conversion of stereotactic coordinates to DICOM coordinates involves a rounding in the longitudinal coordinate to the nearest CT slice, resulting in a potential error in isocentre position of up to half a CT slice thickness. At our institution, we use 1 mm slices for SRS planning, so the maximal error is 0.5 mm. This is small, but represents a systematic deviation that would otherwise go undetected. Institutions using larger CT slices would yield even greater errors. To mitigate this error, we were able to use a function within the TPS that can calculate a stereotactic coordinate from a user-entered DICOM coordinate using the stereotactic localization for that patient. We therefore verify that the DICOM coordinate being sent to the R&V system corresponds to the planned isocentre in stereotactic coordinates. Clinically, we almost always have to modify the exported longitudinal DICOM coordinate to recapitulate the true planned stereotactic coordinate. Multiple independent checks and communication is vital for this process. Discussion The present paper reports results of QA tests during commissioning and routine usage of a commercial FDA-approved SRS system. The software errors and bugs may not get addressed by the manufacturer since the ERGO++ TPS will be discontinued. However, the issues that were discovered in the present TPS should still be assessed and evaluated for any given TPS. In the meanwhile, the vendor has agreed to provide a small patch that would change the way the beams are labelled for easier identification, but each isocentre coordinate and prescription would still have to be manually entered and second-checked. Although the TPS may be replaced, the hardware (i.e., stereotactic cone assembly) will continue to be used and therefore

Errors in Commercial Radiosurgery System 605 represents a more long-term problem. As demonstrated in Figure 3, the cone axis has been shifted from the collimator mechanical axis by 0.5 mm (i.e., a measured displacement of 1 mm) in two clinical days. If all of the mechanical axes should align within 1 mm as per AAPM guidelines, the cone alignment with the collimator alone should be within 0.5 mm, and this is also the recommended tolerance of the CAPCA guidelines (3). The misalignment could be due to an accidental adjustment of the micrometers and/or due to the tray not being pinned or located to the linac head. We are already in communication with the vendor about potential redesigns of the current hardware. For the meanwhile, we have implemented daily EPID-based QA, which is performed immediately prior to the patient coming in for treatment. Conflict of Interest All authors certify that this manuscript has not been published in whole or in part nor is it being considered for publication elsewhere. Dr. Sahgal has received honoraria for educational seminars from Elekta AB, otherwise, the authors have no conflicts of interest to declare. References 1. Schell, M. C., Bova, F. J., Larson, D., Leavitt, D. D., Lutz, W., Podgorsak, E. B., Wu, A. Stereotactic Radiosurgery. American Association of Physicists in Medicine (AAPM) Report 54 (1995). 2. Seung, S. K., Larson, D. A., Galvin, J. M., Mehta, M. P., Potters, L., Schultz, C. J., Yajnik, S. V., Hartford, A. C., Rosenthal, S. A. American college of radiology (ACR) and American society for radiation oncology (ASTRO) practice guideline for the performance of stereotactic radiosurgery (SRS). American Journal of Clinical Oncology 36, 310-315 (2013). DOI: 10.1097/COC.0b013e31826e053d 3. Arsenault, C., Bissonnette, J. P., Dunscombe, P., Gallet, J., Mawko, G., Seuntjens, J. P. Canadian association of provincial cancer agencies (CAPCA): stereotactic radiosurgery/radiotherapy. Standards for Quality Control at Canadian Radiation Treatment Centres (2006). 4. Bissonnette, J. P., Balter, P. A., Dong, L., Langen, K. M., Lovelock, D. M., Miften, M., Moseley, D. J., Pouliot, J., Sonke, J. J., Yoo, S. Quality assurance for image-guided radiation therapy utilizing CTbased technologies: a report of the AAPM TG-179. Medical Physics 39, 1946-1963 (2012). DOI: 10.1118/1.3690466 Received: February 7, 2014; Accepted: February 26, 2014