- Email: [email protected]
Electromagnetic Real-Time Tumor Position Monitoring and Dynamic Multileaf Collimator Tracking Using a Siemens 160 MLC: Geometric and Dosimetric Accuracy of an Integrated System
Int. J. Radiation Oncology Biol. Phys., Vol. 79, No. 2, pp. 579–587, 2011 Copyright 2011 Elsevier Inc. Printed in the USA. All rights reserved 0360-3016/$–see front matter
doi:10.1016/j.ijrobp.2010.03.043
PHYSICS CONTRIBUTION
ELECTROMAGNETIC REAL-TIME TUMOR POSITION MONITORING AND DYNAMIC MULTILEAF COLLIMATOR TRACKING USING A SIEMENS 160 MLC: GEOMETRIC AND DOSIMETRIC ACCURACY OF AN INTEGRATED SYSTEM ANDREAS KRAUSS, M.S.,* SIMEON NILL, PH.D.,* MARTIN TACKE, PH.D.,* AND UWE OELFKE, PH.D.* *Department of Medical Physics in Radiation Oncology, German Cancer Research Center (DKFZ), Heidelberg, Germany Purpose: Dynamic multileaf collimator tracking represents a promising method for high-precision radiotherapy to moving tumors. In the present study, we report on the integration of electromagnetic real-time tumor position monitoring into a multileaf collimator-based tracking system. Methods and Materials: The integrated system was characterized in terms of its geometric and radiologic accuracy. The former was assessed from portal images acquired during radiation delivery to a phantom in tracking mode. The tracking errors were calculated from the positions of the tracking field and of the phantom as extracted from the portal images. Radiologic accuracy was evaluated from film dosimetry performed for conformal and intensity-modulated radiotherapy applied to different phantoms moving on sinusoidal trajectories. A static radiation delivery to the nonmoving target served as a reference for the delivery to the moving phantom with and without tracking applied. Results: Submillimeter tracking accuracy was observed for two-dimensional target motion despite the relatively large system latency of 500 ms. Film dosimetry yielded almost complete recovery of a circular dose distribution with tracking in two dimensions applied: 2%/2 mm gamma-failure rates could be reduced from 59.7% to 3.3%. For single-beam intensity-modulated radiotherapy delivery, accuracy was limited by the finite leaf width. A 2%/2 mm gamma-failure rate of 15.6% remained with tracking applied. Conclusion: The integrated system we have presented marks a major step toward the clinical implementation of high-precision dynamic multileaf collimator tracking. However, several challenges such as irregular motion traces or a thorough quality assurance still need to be addressed. 2011 Elsevier Inc. Adaptive radiotherapy, Intrafractional motion, Real-time, Dynamic multileaf collimator, Electromagnetic tracking.
INTRODUCTION
the compensation of intrafractional organ motion is a more challenging task. The management of intrafractional organ motion consists of two distinct tasks: monitoring the target motion in real time and adapting the dose delivery process to the reported target motion. Techniques to accomplish the former have used external surrogates, implanted internal fiducial markers, or internal anatomic structures. Because of the imperfect correlation between external surrogates and the internal anatomy (5, 6), internal markers are more reliable. Monitoring internal fiducial markers using continuous x-ray imaging, however, results in an increased patient dose (7). A promising alternative is the four-dimensional position monitoring system using implanted electromagnetic transponders developed by Calypso Medical Technologies (Seattle, WA). The Calypso
The goal of modern radiotherapy (RT) is to deliver a lethal amount of dose to cancerous target volumes while sparing the surrounding tissue. Advanced techniques such as intensity-modulated RT (IMRT) allow the delivery of highly conformal dose distributions to static targets. However, interand intrafractional organ motion can compromise the benefits of IMRT. In particular, the dose-blurring effects observed when target motion is present undermine the delivery of optimized dose distributions exhibiting steep dose gradients to protect, for instance, organs at risk (1). Adaptive RT aims to compensate for these motion effects and to restore the highdose conformity of static IMRT deliveries. Although interfractional target motion can be accounted for using recent developments in image-guided RT (2–4), Reprint requests to: Andreas Krauss, M.S., Department of Medical Physics in Radiation Oncology, German Cancer Research Center (DKFZ), Im Neuenheimer Feld 280, Heidelberg 69120 Germany. Tel: (+49) 6221-422418; Fax: (+49) 6221-422572; E-mail: [email protected] Partly supported by Siemens Medical Solutions OCS and Calypso Medical Technologies, Inc.
Conflict of interest: none. Acknowledgments—We thank Dr. Andreas Rau for his support in operating the Calypso System. We also thank Peter Haering and Bernhard Rhein for assistance with the dosimetric evaluation. Received Dec 2, 2009, and in revised form March 16, 2010. Accepted for publication March 17, 2010. 579
I. J. Radiation Oncology d Biology d Physics
580
Volume 79, Number 2, 2011
System provides highly accurate localization of the target identified by three electromagnetic transponders implanted into the target volume without delivering additional ionizing radiation to the patient (8). For the second part of the intrafractional motion management process (i.e., the dose adaption), radiation delivery techniques such as moving the x-ray source (9, 10), moving the treatment couch (11, 12), gating the treatment beam (13, 14), or a gimbals tracking system (15) have been developed. The approach followed in the present study was the compensation of the target motion by a real-time adaption of the aperture of a dynamic multileaf collimator (MLC) (16, 17). We have recently developed a dynamic tumor tracking control (DTTC) system capable of adapting the aperture of a Siemens 160 MLC in real time to the target position provided by a suitable monitoring system. Considerable improvements in dosimetric accuracy for conformal RT and IMRT fields applied to targets moving in two dimensions could be shown, for which the target motion was recorded by linear potentiometers (18). In the present study, we report on the integration of the Calypso System into the DTTC. The performance of the integrated system is characterized in terms of its geometric and dosimetric accuracy. The evaluation was based on conformal RT and IMRT radiation fields applied to different phantoms moving on sinusoidal trajectories in two dimensions. METHODS AND MATERIALS Experimental setup To characterize the performance of our tracking system, we performed a series of phantom experiments to analyze the following three distinct aspects: the overall latency of the integrated system; the geometric tracking accuracy; and the radiologic benefits of our tracking approach compared with standard static radiation deliveries. Figure 1 shows a photograph of the overall experimental setup. The phantom, which had the three electromagnetic transponders embedded, was fixed to the treatment table. The transponder motion was recorded by the electromagnetic array of the Calypso System. The research version of the Calypso System used in the present provided real-time output of the target position by way of an Ethernet connection, with an update rate of approximately 25 Hz to the DTTC outside the treatment room. The DTTC combined the realtime target information with a continuous data feedback loop from the dose delivery process to optimally adapt the leaf positions of a Siemens 160 MLC mounted on a research Siemens ARTISTE linear accelerator to the observed target motion (18). The megavoltage portal imaging device of the linear accelerator (1,024 1,024 pixels, 0.4 mm 0.4 mm pixel size) was used to visually survey proper system performance in the control room. In addition, the recorded frames were used to evaluate both the overall latency and the geometric accuracy of the tracking system. The following coordinate system was used throughout the present study. The origin coincided with the isocenter. The x-axis pointed in the right–left direction for a patient in the head-first, supine position, the y-axis pointed in inferosuperior direction toward the linear accelerator, and the z-axis pointed upward in the posteroanterior direction. The collimator angle was set to 90 such that the leaf travel direction coincided with the y-axis.
Fig. 1. Photograph of experimental setup. Electromagnetic array of Calypso System placed above lung phantom. Linac = linear accelerator; EPID = electronic portal imaging device.
Prediction of future target positions The execution of the individual steps performed by the integrated tracking system required a finite amount of time. Neglecting these would cause the MLC to always lag behind the actual target motion. Therefore, the interval between an initial target movement and the first response of the MLC (i.e., the system’s latency), must be taken properly into account. To compensate for this latency, we implemented an improved prediction algorithm using the method of linear autoregression. The algorithm predicted the future target position as a linear combination of a set of previous target positions. The k-step forward prediction xt+k of the target position at the time t can thus be written as follows: xtþk ¼ a1 xtþ1 þ . þ an xtþn where the {xtþi ji˛½1; n} represent the n latest target points reported to the DTTC. The optimal coefficients ai of the prediction filter were determined by minimizing the mean squared error of the prediction on a training data set. We applied a quadratic fit to the history of target points used for prediction to reduce the signal noise.
Assessment of system latency To quantify the system’s latency, its compensation by the DTTC was switched off and the resulting time lag was determined. Because the aforementioned prediction algorithm not only compensated for the system’s latency, but also for the update interval of 100 ms of
MLC tracking using electromagnetic target position monitoring d A. KRAUSS et al.
the collimator control unit, the prediction time was not set to 0 but was lowered to 100 ms. The phantom was moved in leaf travel direction on a sinusoidal trajectory with an amplitude of 2 cm and a periodicity of 4.5 s. The gantry angle was fixed at 0 , and a circular MLC aperture with 5 cm diameter in the isocenter plane was chosen. During radiation delivery in the tracking mode, portal images were acquired with an update rate of 15 Hz. A metal ball, visible on the portal images, was fixed in the center of the phantom. The centroids of both the circular field aperture and the metal ball in the portal images were determined automatically. An image of the circular radiation field, including the shadow of the metal ball, is shown in Fig. 2. The resulting sampled trajectories of both the field aperture and the target were plotted, and the sine-functions were fitted to it. Thus, the latency could be read off the phase difference of the two fitted curves.
Geometric accuracy To assess the geometric accuracy of the tracking system, the phantom was moved on a Lissajous curve consisting of two sine waves parallel and perpendicular to the leaf travel direction with a periodicity of 4.5 s and 6.5 s. Again, portal images were acquired continuously during radiation delivery in the target tracking mode. Tracking errors were calculated from the distance between the centroids of the radiation field and the metal ball derived from the portal images.
Radiologic accuracy For the assessment of the radiologic benefits achievable with the tracking system, radiochromic films (GAFCHROMIC EBT) placed in between the solid water slices of the phantoms were irradiated. For each set of measurements, three scenarios were investigated: radiation delivery in static MLC mode to the nonmoving phantom defining the reference standard for the tracking experiments, delivery in static MLC mode to the moving target, and delivery in the dynamic MLC tracking mode. The films were scanned and normalized to the maximal dose of the reference standard measurement for each set of experiments. A comparison to this reference measurement was performed in terms of the gamma-test, which brings the dose difference and distance-to-agreement maps together (19). For the calculation of the gamma-metric, we used the maximal value of the reference
Fig. 2. Image of circular field shape, including shadow of metal ball. Green cross lines indicate centroid of radiation field; red square highlights metal ball. In contrast-enhanced display detail, metal ball can be clearly distinguished.
581
dose distribution as the reference value for the accepted percentage of dose deviation. We assumed that the static radiation delivery to the nonmoving target represented optimal delivery in terms of target coverage and healthy tissue sparing and, therefore, considered the gamma-index evaluation (i.e., the estimation of the similarity of a dose distribution to this reference standard distribution) as an appropriate measure of dosimetric accuracy. Advancing from a technical to a more clinically relevant viewpoint, three kinds of experiments using two phantoms were performed: 1. An open, circular radiation field 5 cm in diameter in the isocenter was applied to a simple phantom, which consisted of a stack of water equivalent slices. The phantom was moved both in a twodimensional Lissajous curve and in one-dimensional sinusoidal trajectories parallel and perpendicular to the leaf travel direction with a periodicity of 4.5 s and 6.5 s and amplitude of 2 cm and 1 cm, respectively. This allowed the examination of the influence of the finite leaf width to the total delivery accuracy, which only played a role for the compensation of movements perpendicular to the leaf travel direction. 2. A single beam out of a step-and-shoot IMRT plan (see below) was applied to the same simple phantom in two-dimensional tracking mode. Again, the phantom was moved on the aforementioned Lissajous curve. 3. A complete step-and-shoot IMRT plan was applied to a phantom, which mimicked a human thorax. It consisted of solid water slices of 10 mm thickness with lung equivalent inserts and a tumor inlay. Additional details about the phantom have been previously published (20). The transponders were embedded in the tumor inlay. The phantom was moved on a cos4(t)-shaped trajectory (21) in y-direction with an amplitude of 2.4 cm. The slices of the phantom are aligned perpendicular to the motion axis of the phantom. The delivery was performed in a completely automatic mode using the same communication routines between the Siemens Syngo RT Therapist workspace and the linear accelerator control console, just as they are used in clinical practice. The standard five-beam treatment plan was generated for a computed tomography scan of the lung phantom.
RESULTS Latency Figure 3a shows the results of the latency measurement. The crosses indicate the trajectories of the field aperture and the metal ball as extracted from the portal images. The sine curves fitted to the experimental data and the resulting fit parameters are also shown. Of the phase difference of the two sine curves, an overall system latency of 500 ms was calculated. For all additional experiments, this value was used as the prediction interval. The overall latency of the tracking system was composed of the processing times of the Calypso System and the MLC control unit. The Calypso software calculated its own latency and reported it to the DTTC, together with the position update. Figure 3b displays the rather broad distribution of the Calypso processing times, with a mean value of 109 ms. The cause of the variation in the Calypso latency values has been previously discussed in detail (22). The overall system latency was dominated by the processing time of the MLC control unit.
582
I. J. Radiation Oncology d Biology d Physics
Volume 79, Number 2, 2011
Fig. 3. (a) Trajectories of circular field’s centroid (red) and target position (blue). Crosses indicate values extracted from portal images. Solid lines represent sine functions fitted to measured data. (b) Internal latency of Calypso System.
Geometric accuracy In Fig. 4, the trajectories of the centroid of the field aperture and the target position were plotted, together with the tracking error (i.e., the difference between these two curves). In both cases, the absolute of the tracking errors remained at <1.6 mm. A root mean squared error of 0.69 mm and 0.80 mm was observed for the directions parallel and perpendicular to the leaf travel direction, respectively. Parallel to the leaf travel direction, the maximal tracking errors occurred in the regions of the greatest leaf velocities. Perpendicular to the leaf travel direction, the greatest deviations were observed in the peaks of the trajectory. Radiologic accuracy The upper row of Fig. 5 shows the dose distributions for the circular field applied to the nonmoving target in the static MLC mode, the target moving in two directions in the static MLC mode, and the moving target in the dynamic tracking MLC mode. For the no-tracking case, strong blurring of the static reference dose distribution was observed. The high-dose area inside the 95% isodose line decreased from
13.8 to 8.1 cm2. In contrast, the region between the 20% and 95% isodose lines increased from 11.7 to 24.6 cm2. Except for a smoothing of the sharp edges, the reference dose distribution could be well restored in the tracking case. The lower row of Fig. 5 shows maps of the points failing the 2%/2 mm gamma-criterion (i.e., having a gamma-value >1). A large ring of points receiving an underdosage (blue spots) inside the target volume followed by an overdosage region (red spots) outside the target volume could be observed for the no-tracking case. In contrast, only very few points near the closing position of the adjacent leaves failed the gamma-test in the tracking case. Table 1 lists the gamma-test results for tracking in two dimensions and for tracking along the directions parallel and perpendicular to the leaf travel direction. For tracking parallel to the leaf travel direction, an almost complete recovery of the reference dose distribution was observed. Only 0.6% of the evaluated points failed the 2%/2-mm gamma-criterion. For tracking perpendicular to the leaf travel direction, an increased failure rate of 3.6% was observed. For target displacements perpendicular to the leaf travel direction, the
Fig. 4. Tracking accuracy in direction parallel (a) and perpendicular (b) to leaf travel direction. Trajectories of both centroid of radiation field and target position and difference between these curves displayed.
MLC tracking using electromagnetic target position monitoring d A. KRAUSS et al.
583
Fig. 5. Dose distributions of circular field applied to target moving in two directions (Upper row). Gamma-test results displayed as grayscale dose distribution with gamma-indexes > 1 marked in red for overdosage and blue for underdosage (Lower row).
planned field shape could not be completely restored because of the finite leaf width of 5 mm, which was therefore the main contribution to the radiologic inaccuracy for twodimensional tracking in this experiment. Figure 6 shows the results of the film evaluation of a single IMRT beam applied to a target moving in two dimensions. Again, dose distributions (upper row) and maps of points failing the 2%/2 mm gamma-criterion (lower row) are displayed. For the tracking case, an improvement of the strong blurring effect of the no-tracking case was clearly visible. The gamma failure rate decreased from 55% to 15.6% and from 31% to 7.6% for the 2%/2 mm and 3%/3 mm criterion, respectively. However, an underdosage of the sharp target edges and an overdosage of the surrounding areas above and below the target area remained when tracking was applied.
In a third set of measurements, the radiologic effect of our tracking approach to a complete IMRT plan for a lung tumor site was assessed. Four films were sandwiched between the slices of the lung phantom. Two films (Films 2 and 3) were placed near the center of the phantom’s tumor inlay; Film 1 was placed 5 mm inside the target volume and Film 4 was placed 5 mm outside the target volume. Although for the two central sections (No. 2 and 3), the target motion induced only minor discrepancies between planned and delivered dose, the planned dose distribution to Films 1 and 4 was completely compromised. The 2%/2 mm gamma-analysis results are listed in Table 2. Figure 7 shows the horizontal dose profiles of Films 1 and 4. Both the strong underdosage of Film 1 and the overdosage of Film 4 observed in the no-tracking case were substantially reduced using the tracking technique. Tracking resulted in
Table 1. Failure rates of gamma-test for circular field applied to different target motion patterns Without tracking
With tracking
Target motion
2%/2 mm (%)
3%/3 mm (%)
2%/2 mm (%)
3%/3 mm (%)
Two-dimensional Parallel to leaf travel direction Perpendicular to leaf travel direction
59.7 51.1 14.8
40.6 32.3 1.5
3.3 0.6 3.6
0.5 0.1 0.9
I. J. Radiation Oncology d Biology d Physics
584
Volume 79, Number 2, 2011
Fig. 6. Dose distributions of single intensity-modulated radiotherapy (IMRT) beam applied to target moving in two directions (Upper row). Gamma-test results displayed as grayscale dose distribution with gamma-indexes >1 marked in red for overdosage and blue for underdosage (Lower row).
of our current version of the tracking system was large, but efforts will be made to reduce the processing time of the collimator control unit. The experiments performed in the present study represented the ideal case of perfectly regular target motion patterns. The linear filter used in our study to predict the sinusoidal motion patterns, therefore, induced almost no error to the overall system, despite the large system latency. However, for irregular, patient-specific breathing motions, the predictor’s performance is likely to be severely reduced. Future work will therefore include the implementation and testing of more advanced prediction filters (24, 25). Film dosimetry yielded considerable improvements for all the investigated scenarios. An almost complete recovery of the planned dose distribution was observed for open field tracking. A 2%/2 mm gamma-failure rate of 0.6% and 3.6% for target motion parallel and perpendicular to the leaf travel direction, respectively, was observed. These findings emphasize that the prediction errors play no role for the
a reduction of the mean overdosage of Film 4 (i.e., the mean dose difference between the measurement and the reference standard scan), from 20.9% to 5.2% of the maximal target dose. DISCUSSION The results of the experiments presented in the present study have demonstrated the feasibility of intrafractional motion compensation using a Siemens 160 MLC by monitoring the tumor motion using the Calypso System. With a geometric tracking accuracy of <1 mm for a target moving in two dimensions on sinusoidal trajectories, our system can compete with similar systems (22). For the system’s latency, a value of 500 ms was observed, which mainly originated from the collimator control unit. Compared with other real-time tracking solutions, such as the CyberKnife robotic treatment device (Accuray, Sunnyvale, CA) or the combination of a Millennium MLC (Varian Medical Systems, Palo Alto, CA) and the Calypso System with a latency of 115 ms and 220 ms (22, 23), the latency
Table 2. Failure rates of gamma-test for five-beam intensity-modulated radiotherapy plan applied to lung phantom moving on onedimensional cos4(t) trajectory with amplitude of 24 mm and periodicity of 5 s Without tracking Film position 1 2 3 4
With tracking
2%/2 mm (%)
3%/3 mm (%)
2%/2 mm (%)
3%/3 mm (%)
99.6 10.3 27.6 98.6
98.5 2.4 11.1 96.7
8.6 0.3 1.8 65.1
2.2 0.0 0.1 51.8
For no-tracking case, treatment isocenter placed in full exhale phase of target trajectory. Calculation of gamma-metric refers to same normalization value for four films (i.e., maximal dose recorded for reference measurement).
MLC tracking using electromagnetic target position monitoring d A. KRAUSS et al.
585
Fig. 7. (a,b) Gamma-test results for Films 1 and 4 displayed as dose distributions recorded on tracking film, with points failing 2%/2 mm gamma-criterion marked in red and blue for overdosage and underdosage, respectively. (c,d) Horizontal profiles of films 1 and 4 for reference, tracking, and no-tracking cases. Dose values normalized to maximal dose delivered to tumor.
regular motion traces used in our study. The tracking accuracy was primarily limited by the finite leaf width of 5 mm in the isocenter. Assuming that target motion was taken into account using appropriate margins for the three investigated scenarios, we considered the region inside the 95% isodose line of the dose distributions as the clinical target volume (CTV). Target motion resulted in a 41% decrease of the CTV. The region outside the CTV receiving #20% of the target dose was increased by 110%. Both effects were negligible in the tracking case (CTV decrease of 1%, irradiated healthy tissue increase of 5%). For a single IMRT beam applied to the phantom moving in two dimensions, the 2%/2-mm gamma-failure rate was reduced from 54.6% to 15.6% by applying the tracking technique. Again, the remaining inaccuracy could be addressed to the finite leaf width. Because several segments of the IMRT field exhibited large leaf apertures directly adjacent to a closed leaf pair, the effect of the finite leaf width on the IMRT dose distribution was by far more pronounced than
for the circular field. We will investigate whether an optimized collimator angle can help to generate field shapes that will be better suited for delivery in two-dimensional tracking mode. To investigate the clinical benefits of the tracking system, the dosimetric accuracy was assessed for a five-beam IMRT plan. The strong underdosage of the tumor edge traveling out of the treatment field without tracking applied could be compensated effectively. The 2%/2 mm gamma-failure rate decreased from 99.6% to 8.6%. In clinical practice, this underdosage would have been addressed by adding appropriate margins in the planning phase. The overdosage of the surrounding tissue traveling into the treatment beam without tracking recorded on Film 4 showed that the dose gradients intended to protect the healthy tissue from an excessive dose were severely compromised without tracking. This effect could not be completely eliminated. Nevertheless, a reduction of the 3%/3 mm gamma-failure rate from 96.7% to 51.8% and of the mean overdosage from 20.9% to 5.2% of the target dose observed for a film placed 5 mm outside the target volume showed at least considerable
586
I. J. Radiation Oncology d Biology d Physics
improvement when the tracking technique was applied. The dose profiles displayed in Fig. 7d show that the maximal dose recorded on that particular film for the reference measurement was approximately 55% of the maximal dose delivered to the tumor. This indicates that the film was placed in a region of steep dose gradients. The remaining inaccuracy for the dose delivery thus showed the sensitivity of the steep dose gradients achievable with IMRT to minor target displacements. Even for the idealized conditions of this tracking experiment, we were not able to completely restore the planned dose distribution. We are planning to perform experiments similar to those presented in the present study with phantoms moving on irregular, patient-specific motion traces. We are, therefore, currently incorporating a three-dimensional programmable motion stage into our experimental setup. Recently, the implantation of the Beacon transponders into the prostate was approved in the European Union. We aim to bring the integrated tracking system presented in the present work into clinical prostate treatment routines. Toward this goal, the thorough development of adequate quality assurance routines is essential. For lung applications, a novel transponder design is currently under development. A recent study showed 100% long-term fixation rates for these transponders implanted bronchoscopically into canine lungs (26).
Volume 79, Number 2, 2011
Another challenge to be addressed in the future is the management of more complex target motion. In the current setup, the integrated tracking system compensated in real time for three-dimensional rigid target translations according to the position information of the centroid of the three transponders. A continuous update of the individual transponder positions could, however, yield insights into possible tumor rotations or deformations. In addition, we have previously reported on synchronized target position monitoring with the Calypso System and fluoroscopic x-ray imaging (27). The combination of these techniques could allow the monitoring of target deformations and rotations and the motion of organs at risk. Future work will include strategies to realign the MLC leaves in real time to also account for this additional target information.
CONCLUSION The integrated system characterized in the present study represents a major step toward the clinical implementation of intrafractional tumor motion compensation. We could demonstrate highly accurate target tracking and showed that even complex IMRT dose distributions can be recovered to a high degree using the integrated tracking system. However, several remaining challenges need to be addressed.
REFERENCES 1. Bortfeld T, Jiang SB, Rietzel E. Effects of motion on the total dose distribution. Semin Radiat Oncol 2004;14:41–51. 2. Xing L, Thorndyke B, Schreibmann E, et al. Overview of image-guided radiation therapy. Med Dosim 2006;31: 91–112. 3. Oelfke U, Tucking T, Nill S, et al. Linac-integrated kV-cone beam CT: Technical features and first applications. Med Dosim 2006;31:62–70. 4. Sonke JJ, Rossi M, Wolthaus J, et al. Frameless stereotactic body radiotherapy for lung cancer using four-dimensional cone beam CT guidance. Int J Radiat Oncol Biol Phys 2009; 74:567–574. 5. Gierga DP, Brewer J, Sharp GC, et al. The correlation between internal and external markers for abdominal tumors: Implications for respiratory gating. Int J Radiat Oncol Biol Phys 2005;61:1551–1558. 6. Beddar AS, Kainz K, Briere TM, et al. Correlation between internal fiducial tumor motion and external marker motion for liver tumors imaged with 4D-CT. Int J Radiat Oncol Biol Phys 2007;67:630–638. 7. Shirato H, Oita M, Fujita K, et al. Feasibility of synchronization of realtime tumor-tracking radiotherapy and intensitymodulated radiotherapy from viewpoint of excessive dose from fluoroscopy. Int J Radiat Oncol Biol Phys 2004;60: 334–341. 8. Balter JM, Wright JN, Newell LJ, et al. Accuracy of a wireless localization system for radiotherapy. Int J Radiat Oncol Biol Phys 2005;61:933–937. 9. Schweikard A, Shiomi H, Adler J. Respiration tracking in radiosurgery. Med Phys 2004;31:2738–2741. 10. Seppenwoolde Y, Berbeco RI, Nishioka S, et al. Accuracy of tumor motion compensation algorithm from a robotic respira-
11. 12.
13. 14. 15. 16. 17. 18. 19. 20.
tory tracking system: A simulation study. Med Phys 2007;34: 2774–2784. D’Souza WD, McAvoy TJ. An analysis of the treatment couch and control system dynamics for respiration-induced motion compensation. Med Phys 2006;33:4701–4709. Qiu P, D’Souza WD, McAvoy TJ, et al. Inferential modelling and predictive feedback control in real-time motion compensation using the treatment couch during radiotherapy. Phys Med Biol 2007;52:5831–5854. Shirato H, Shimizu S, Kunieda T, et al. Physical aspects of a real-time tumor-tracking system for gated radiotherapy. Int J Radiat Oncol Biol Phys 2000;48:1187–1195. Shen S, Duan J, Fiveash JB, et al. Validation of target volume and position in respiratory gated CT planning and treatment. Med Phys 2003;30:3196–3205. Takayama K, Mizowaki T, Kokubo M, et al. Initial validations for pursuing irradiation using a gimbals tracking system. Radiother Oncol 2009;93:45–49. McQuaid D, Partridge M, Symonds-Tayler JR, et al. Targettracking deliveries on an Elekta linac: A feasibility study. Phys Med Biol 2009;54:3563–3578. Sawant A, Venkat R, Srivastava V, et al. Management of three dimensional intrafraction motion through real-time DMLC tracking. Med Phys 2008;35:2050–2061. Tacke MB, Nill S, Krauss A, et al. Real-time tumor tracking: Automatic compensation of target motion using the Siemens 160 MLC. Med Phys 2010;37:753–761. Low DA, Dempsey JF. Evaluation of the gamma dose distribution comparison method. Med Phys 2003;30:2455–2464. Dietrich L, Tucking T, Nill S, et al. Compensation for respiratory motion by gated radiotherapy: An experimental study. Phys Med Biol 2005;50:2405–2414.
MLC tracking using electromagnetic target position monitoring d A. KRAUSS et al.
21. Lujan AE, Larsen EW, Balter JM, et al. A method for incorporating organ motion due to breathing into 3D dose calculations. Med Phys 1999;26:715–720. 22. Sawant A, Smith RL, Venkat RB, et al. Toward submillimeter accuracy in the management of intrafraction motion: The integration of real-time internal position monitoring and multileaf collimator target tracking. Int J Radiat Oncol Biol Phys 2009; 74:575–582. 23. Hoogeman M, Prevost JB, Nuyttens J, et al. Clinical accuracy of the respiratory tumor tracking system of the CyberKnife: Assessment by analysis of log files. Int J Radiat Oncol Biol Phys 2009;74:297–303.
587
24. Murphy MJ, Dieterich S. Comparative performance of linear and nonlinear neural networks to predict irregular breathing. Phys Med Biol 2006;51:5903–5914. 25. Ruan D. Kernel density estimation-based real-time prediction for respiratory motion. Phys Med Biol 2010;55:1311–1326. 26. Mayse ML, Smith RL, Park M, et al. Development of a nonmigrating electromagnetic transponder system for lung tumor tracking. Int J Radiat Oncol Biol Phys 2008;72:S430. 27. Rau AW, Nill S, Eidens RS, et al. Synchronized tumour tracking with electromagnetic transponders and kV x-ray imaging: Evaluation based on a thorax phantom. Phys Med Biol 2008; 53:3789–3805.
doi:10.1016/j.ijrobp.2010.03.043
PHYSICS CONTRIBUTION
ELECTROMAGNETIC REAL-TIME TUMOR POSITION MONITORING AND DYNAMIC MULTILEAF COLLIMATOR TRACKING USING A SIEMENS 160 MLC: GEOMETRIC AND DOSIMETRIC ACCURACY OF AN INTEGRATED SYSTEM ANDREAS KRAUSS, M.S.,* SIMEON NILL, PH.D.,* MARTIN TACKE, PH.D.,* AND UWE OELFKE, PH.D.* *Department of Medical Physics in Radiation Oncology, German Cancer Research Center (DKFZ), Heidelberg, Germany Purpose: Dynamic multileaf collimator tracking represents a promising method for high-precision radiotherapy to moving tumors. In the present study, we report on the integration of electromagnetic real-time tumor position monitoring into a multileaf collimator-based tracking system. Methods and Materials: The integrated system was characterized in terms of its geometric and radiologic accuracy. The former was assessed from portal images acquired during radiation delivery to a phantom in tracking mode. The tracking errors were calculated from the positions of the tracking field and of the phantom as extracted from the portal images. Radiologic accuracy was evaluated from film dosimetry performed for conformal and intensity-modulated radiotherapy applied to different phantoms moving on sinusoidal trajectories. A static radiation delivery to the nonmoving target served as a reference for the delivery to the moving phantom with and without tracking applied. Results: Submillimeter tracking accuracy was observed for two-dimensional target motion despite the relatively large system latency of 500 ms. Film dosimetry yielded almost complete recovery of a circular dose distribution with tracking in two dimensions applied: 2%/2 mm gamma-failure rates could be reduced from 59.7% to 3.3%. For single-beam intensity-modulated radiotherapy delivery, accuracy was limited by the finite leaf width. A 2%/2 mm gamma-failure rate of 15.6% remained with tracking applied. Conclusion: The integrated system we have presented marks a major step toward the clinical implementation of high-precision dynamic multileaf collimator tracking. However, several challenges such as irregular motion traces or a thorough quality assurance still need to be addressed. 2011 Elsevier Inc. Adaptive radiotherapy, Intrafractional motion, Real-time, Dynamic multileaf collimator, Electromagnetic tracking.
INTRODUCTION
the compensation of intrafractional organ motion is a more challenging task. The management of intrafractional organ motion consists of two distinct tasks: monitoring the target motion in real time and adapting the dose delivery process to the reported target motion. Techniques to accomplish the former have used external surrogates, implanted internal fiducial markers, or internal anatomic structures. Because of the imperfect correlation between external surrogates and the internal anatomy (5, 6), internal markers are more reliable. Monitoring internal fiducial markers using continuous x-ray imaging, however, results in an increased patient dose (7). A promising alternative is the four-dimensional position monitoring system using implanted electromagnetic transponders developed by Calypso Medical Technologies (Seattle, WA). The Calypso
The goal of modern radiotherapy (RT) is to deliver a lethal amount of dose to cancerous target volumes while sparing the surrounding tissue. Advanced techniques such as intensity-modulated RT (IMRT) allow the delivery of highly conformal dose distributions to static targets. However, interand intrafractional organ motion can compromise the benefits of IMRT. In particular, the dose-blurring effects observed when target motion is present undermine the delivery of optimized dose distributions exhibiting steep dose gradients to protect, for instance, organs at risk (1). Adaptive RT aims to compensate for these motion effects and to restore the highdose conformity of static IMRT deliveries. Although interfractional target motion can be accounted for using recent developments in image-guided RT (2–4), Reprint requests to: Andreas Krauss, M.S., Department of Medical Physics in Radiation Oncology, German Cancer Research Center (DKFZ), Im Neuenheimer Feld 280, Heidelberg 69120 Germany. Tel: (+49) 6221-422418; Fax: (+49) 6221-422572; E-mail: [email protected] Partly supported by Siemens Medical Solutions OCS and Calypso Medical Technologies, Inc.
Conflict of interest: none. Acknowledgments—We thank Dr. Andreas Rau for his support in operating the Calypso System. We also thank Peter Haering and Bernhard Rhein for assistance with the dosimetric evaluation. Received Dec 2, 2009, and in revised form March 16, 2010. Accepted for publication March 17, 2010. 579
I. J. Radiation Oncology d Biology d Physics
580
Volume 79, Number 2, 2011
System provides highly accurate localization of the target identified by three electromagnetic transponders implanted into the target volume without delivering additional ionizing radiation to the patient (8). For the second part of the intrafractional motion management process (i.e., the dose adaption), radiation delivery techniques such as moving the x-ray source (9, 10), moving the treatment couch (11, 12), gating the treatment beam (13, 14), or a gimbals tracking system (15) have been developed. The approach followed in the present study was the compensation of the target motion by a real-time adaption of the aperture of a dynamic multileaf collimator (MLC) (16, 17). We have recently developed a dynamic tumor tracking control (DTTC) system capable of adapting the aperture of a Siemens 160 MLC in real time to the target position provided by a suitable monitoring system. Considerable improvements in dosimetric accuracy for conformal RT and IMRT fields applied to targets moving in two dimensions could be shown, for which the target motion was recorded by linear potentiometers (18). In the present study, we report on the integration of the Calypso System into the DTTC. The performance of the integrated system is characterized in terms of its geometric and dosimetric accuracy. The evaluation was based on conformal RT and IMRT radiation fields applied to different phantoms moving on sinusoidal trajectories in two dimensions. METHODS AND MATERIALS Experimental setup To characterize the performance of our tracking system, we performed a series of phantom experiments to analyze the following three distinct aspects: the overall latency of the integrated system; the geometric tracking accuracy; and the radiologic benefits of our tracking approach compared with standard static radiation deliveries. Figure 1 shows a photograph of the overall experimental setup. The phantom, which had the three electromagnetic transponders embedded, was fixed to the treatment table. The transponder motion was recorded by the electromagnetic array of the Calypso System. The research version of the Calypso System used in the present provided real-time output of the target position by way of an Ethernet connection, with an update rate of approximately 25 Hz to the DTTC outside the treatment room. The DTTC combined the realtime target information with a continuous data feedback loop from the dose delivery process to optimally adapt the leaf positions of a Siemens 160 MLC mounted on a research Siemens ARTISTE linear accelerator to the observed target motion (18). The megavoltage portal imaging device of the linear accelerator (1,024 1,024 pixels, 0.4 mm 0.4 mm pixel size) was used to visually survey proper system performance in the control room. In addition, the recorded frames were used to evaluate both the overall latency and the geometric accuracy of the tracking system. The following coordinate system was used throughout the present study. The origin coincided with the isocenter. The x-axis pointed in the right–left direction for a patient in the head-first, supine position, the y-axis pointed in inferosuperior direction toward the linear accelerator, and the z-axis pointed upward in the posteroanterior direction. The collimator angle was set to 90 such that the leaf travel direction coincided with the y-axis.
Fig. 1. Photograph of experimental setup. Electromagnetic array of Calypso System placed above lung phantom. Linac = linear accelerator; EPID = electronic portal imaging device.
Prediction of future target positions The execution of the individual steps performed by the integrated tracking system required a finite amount of time. Neglecting these would cause the MLC to always lag behind the actual target motion. Therefore, the interval between an initial target movement and the first response of the MLC (i.e., the system’s latency), must be taken properly into account. To compensate for this latency, we implemented an improved prediction algorithm using the method of linear autoregression. The algorithm predicted the future target position as a linear combination of a set of previous target positions. The k-step forward prediction xt+k of the target position at the time t can thus be written as follows: xtþk ¼ a1 xtþ1 þ . þ an xtþn where the {xtþi ji˛½1; n} represent the n latest target points reported to the DTTC. The optimal coefficients ai of the prediction filter were determined by minimizing the mean squared error of the prediction on a training data set. We applied a quadratic fit to the history of target points used for prediction to reduce the signal noise.
Assessment of system latency To quantify the system’s latency, its compensation by the DTTC was switched off and the resulting time lag was determined. Because the aforementioned prediction algorithm not only compensated for the system’s latency, but also for the update interval of 100 ms of
MLC tracking using electromagnetic target position monitoring d A. KRAUSS et al.
the collimator control unit, the prediction time was not set to 0 but was lowered to 100 ms. The phantom was moved in leaf travel direction on a sinusoidal trajectory with an amplitude of 2 cm and a periodicity of 4.5 s. The gantry angle was fixed at 0 , and a circular MLC aperture with 5 cm diameter in the isocenter plane was chosen. During radiation delivery in the tracking mode, portal images were acquired with an update rate of 15 Hz. A metal ball, visible on the portal images, was fixed in the center of the phantom. The centroids of both the circular field aperture and the metal ball in the portal images were determined automatically. An image of the circular radiation field, including the shadow of the metal ball, is shown in Fig. 2. The resulting sampled trajectories of both the field aperture and the target were plotted, and the sine-functions were fitted to it. Thus, the latency could be read off the phase difference of the two fitted curves.
Geometric accuracy To assess the geometric accuracy of the tracking system, the phantom was moved on a Lissajous curve consisting of two sine waves parallel and perpendicular to the leaf travel direction with a periodicity of 4.5 s and 6.5 s. Again, portal images were acquired continuously during radiation delivery in the target tracking mode. Tracking errors were calculated from the distance between the centroids of the radiation field and the metal ball derived from the portal images.
Radiologic accuracy For the assessment of the radiologic benefits achievable with the tracking system, radiochromic films (GAFCHROMIC EBT) placed in between the solid water slices of the phantoms were irradiated. For each set of measurements, three scenarios were investigated: radiation delivery in static MLC mode to the nonmoving phantom defining the reference standard for the tracking experiments, delivery in static MLC mode to the moving target, and delivery in the dynamic MLC tracking mode. The films were scanned and normalized to the maximal dose of the reference standard measurement for each set of experiments. A comparison to this reference measurement was performed in terms of the gamma-test, which brings the dose difference and distance-to-agreement maps together (19). For the calculation of the gamma-metric, we used the maximal value of the reference
Fig. 2. Image of circular field shape, including shadow of metal ball. Green cross lines indicate centroid of radiation field; red square highlights metal ball. In contrast-enhanced display detail, metal ball can be clearly distinguished.
581
dose distribution as the reference value for the accepted percentage of dose deviation. We assumed that the static radiation delivery to the nonmoving target represented optimal delivery in terms of target coverage and healthy tissue sparing and, therefore, considered the gamma-index evaluation (i.e., the estimation of the similarity of a dose distribution to this reference standard distribution) as an appropriate measure of dosimetric accuracy. Advancing from a technical to a more clinically relevant viewpoint, three kinds of experiments using two phantoms were performed: 1. An open, circular radiation field 5 cm in diameter in the isocenter was applied to a simple phantom, which consisted of a stack of water equivalent slices. The phantom was moved both in a twodimensional Lissajous curve and in one-dimensional sinusoidal trajectories parallel and perpendicular to the leaf travel direction with a periodicity of 4.5 s and 6.5 s and amplitude of 2 cm and 1 cm, respectively. This allowed the examination of the influence of the finite leaf width to the total delivery accuracy, which only played a role for the compensation of movements perpendicular to the leaf travel direction. 2. A single beam out of a step-and-shoot IMRT plan (see below) was applied to the same simple phantom in two-dimensional tracking mode. Again, the phantom was moved on the aforementioned Lissajous curve. 3. A complete step-and-shoot IMRT plan was applied to a phantom, which mimicked a human thorax. It consisted of solid water slices of 10 mm thickness with lung equivalent inserts and a tumor inlay. Additional details about the phantom have been previously published (20). The transponders were embedded in the tumor inlay. The phantom was moved on a cos4(t)-shaped trajectory (21) in y-direction with an amplitude of 2.4 cm. The slices of the phantom are aligned perpendicular to the motion axis of the phantom. The delivery was performed in a completely automatic mode using the same communication routines between the Siemens Syngo RT Therapist workspace and the linear accelerator control console, just as they are used in clinical practice. The standard five-beam treatment plan was generated for a computed tomography scan of the lung phantom.
RESULTS Latency Figure 3a shows the results of the latency measurement. The crosses indicate the trajectories of the field aperture and the metal ball as extracted from the portal images. The sine curves fitted to the experimental data and the resulting fit parameters are also shown. Of the phase difference of the two sine curves, an overall system latency of 500 ms was calculated. For all additional experiments, this value was used as the prediction interval. The overall latency of the tracking system was composed of the processing times of the Calypso System and the MLC control unit. The Calypso software calculated its own latency and reported it to the DTTC, together with the position update. Figure 3b displays the rather broad distribution of the Calypso processing times, with a mean value of 109 ms. The cause of the variation in the Calypso latency values has been previously discussed in detail (22). The overall system latency was dominated by the processing time of the MLC control unit.
582
I. J. Radiation Oncology d Biology d Physics
Volume 79, Number 2, 2011
Fig. 3. (a) Trajectories of circular field’s centroid (red) and target position (blue). Crosses indicate values extracted from portal images. Solid lines represent sine functions fitted to measured data. (b) Internal latency of Calypso System.
Geometric accuracy In Fig. 4, the trajectories of the centroid of the field aperture and the target position were plotted, together with the tracking error (i.e., the difference between these two curves). In both cases, the absolute of the tracking errors remained at <1.6 mm. A root mean squared error of 0.69 mm and 0.80 mm was observed for the directions parallel and perpendicular to the leaf travel direction, respectively. Parallel to the leaf travel direction, the maximal tracking errors occurred in the regions of the greatest leaf velocities. Perpendicular to the leaf travel direction, the greatest deviations were observed in the peaks of the trajectory. Radiologic accuracy The upper row of Fig. 5 shows the dose distributions for the circular field applied to the nonmoving target in the static MLC mode, the target moving in two directions in the static MLC mode, and the moving target in the dynamic tracking MLC mode. For the no-tracking case, strong blurring of the static reference dose distribution was observed. The high-dose area inside the 95% isodose line decreased from
13.8 to 8.1 cm2. In contrast, the region between the 20% and 95% isodose lines increased from 11.7 to 24.6 cm2. Except for a smoothing of the sharp edges, the reference dose distribution could be well restored in the tracking case. The lower row of Fig. 5 shows maps of the points failing the 2%/2 mm gamma-criterion (i.e., having a gamma-value >1). A large ring of points receiving an underdosage (blue spots) inside the target volume followed by an overdosage region (red spots) outside the target volume could be observed for the no-tracking case. In contrast, only very few points near the closing position of the adjacent leaves failed the gamma-test in the tracking case. Table 1 lists the gamma-test results for tracking in two dimensions and for tracking along the directions parallel and perpendicular to the leaf travel direction. For tracking parallel to the leaf travel direction, an almost complete recovery of the reference dose distribution was observed. Only 0.6% of the evaluated points failed the 2%/2-mm gamma-criterion. For tracking perpendicular to the leaf travel direction, an increased failure rate of 3.6% was observed. For target displacements perpendicular to the leaf travel direction, the
Fig. 4. Tracking accuracy in direction parallel (a) and perpendicular (b) to leaf travel direction. Trajectories of both centroid of radiation field and target position and difference between these curves displayed.
MLC tracking using electromagnetic target position monitoring d A. KRAUSS et al.
583
Fig. 5. Dose distributions of circular field applied to target moving in two directions (Upper row). Gamma-test results displayed as grayscale dose distribution with gamma-indexes > 1 marked in red for overdosage and blue for underdosage (Lower row).
planned field shape could not be completely restored because of the finite leaf width of 5 mm, which was therefore the main contribution to the radiologic inaccuracy for twodimensional tracking in this experiment. Figure 6 shows the results of the film evaluation of a single IMRT beam applied to a target moving in two dimensions. Again, dose distributions (upper row) and maps of points failing the 2%/2 mm gamma-criterion (lower row) are displayed. For the tracking case, an improvement of the strong blurring effect of the no-tracking case was clearly visible. The gamma failure rate decreased from 55% to 15.6% and from 31% to 7.6% for the 2%/2 mm and 3%/3 mm criterion, respectively. However, an underdosage of the sharp target edges and an overdosage of the surrounding areas above and below the target area remained when tracking was applied.
In a third set of measurements, the radiologic effect of our tracking approach to a complete IMRT plan for a lung tumor site was assessed. Four films were sandwiched between the slices of the lung phantom. Two films (Films 2 and 3) were placed near the center of the phantom’s tumor inlay; Film 1 was placed 5 mm inside the target volume and Film 4 was placed 5 mm outside the target volume. Although for the two central sections (No. 2 and 3), the target motion induced only minor discrepancies between planned and delivered dose, the planned dose distribution to Films 1 and 4 was completely compromised. The 2%/2 mm gamma-analysis results are listed in Table 2. Figure 7 shows the horizontal dose profiles of Films 1 and 4. Both the strong underdosage of Film 1 and the overdosage of Film 4 observed in the no-tracking case were substantially reduced using the tracking technique. Tracking resulted in
Table 1. Failure rates of gamma-test for circular field applied to different target motion patterns Without tracking
With tracking
Target motion
2%/2 mm (%)
3%/3 mm (%)
2%/2 mm (%)
3%/3 mm (%)
Two-dimensional Parallel to leaf travel direction Perpendicular to leaf travel direction
59.7 51.1 14.8
40.6 32.3 1.5
3.3 0.6 3.6
0.5 0.1 0.9
I. J. Radiation Oncology d Biology d Physics
584
Volume 79, Number 2, 2011
Fig. 6. Dose distributions of single intensity-modulated radiotherapy (IMRT) beam applied to target moving in two directions (Upper row). Gamma-test results displayed as grayscale dose distribution with gamma-indexes >1 marked in red for overdosage and blue for underdosage (Lower row).
of our current version of the tracking system was large, but efforts will be made to reduce the processing time of the collimator control unit. The experiments performed in the present study represented the ideal case of perfectly regular target motion patterns. The linear filter used in our study to predict the sinusoidal motion patterns, therefore, induced almost no error to the overall system, despite the large system latency. However, for irregular, patient-specific breathing motions, the predictor’s performance is likely to be severely reduced. Future work will therefore include the implementation and testing of more advanced prediction filters (24, 25). Film dosimetry yielded considerable improvements for all the investigated scenarios. An almost complete recovery of the planned dose distribution was observed for open field tracking. A 2%/2 mm gamma-failure rate of 0.6% and 3.6% for target motion parallel and perpendicular to the leaf travel direction, respectively, was observed. These findings emphasize that the prediction errors play no role for the
a reduction of the mean overdosage of Film 4 (i.e., the mean dose difference between the measurement and the reference standard scan), from 20.9% to 5.2% of the maximal target dose. DISCUSSION The results of the experiments presented in the present study have demonstrated the feasibility of intrafractional motion compensation using a Siemens 160 MLC by monitoring the tumor motion using the Calypso System. With a geometric tracking accuracy of <1 mm for a target moving in two dimensions on sinusoidal trajectories, our system can compete with similar systems (22). For the system’s latency, a value of 500 ms was observed, which mainly originated from the collimator control unit. Compared with other real-time tracking solutions, such as the CyberKnife robotic treatment device (Accuray, Sunnyvale, CA) or the combination of a Millennium MLC (Varian Medical Systems, Palo Alto, CA) and the Calypso System with a latency of 115 ms and 220 ms (22, 23), the latency
Table 2. Failure rates of gamma-test for five-beam intensity-modulated radiotherapy plan applied to lung phantom moving on onedimensional cos4(t) trajectory with amplitude of 24 mm and periodicity of 5 s Without tracking Film position 1 2 3 4
With tracking
2%/2 mm (%)
3%/3 mm (%)
2%/2 mm (%)
3%/3 mm (%)
99.6 10.3 27.6 98.6
98.5 2.4 11.1 96.7
8.6 0.3 1.8 65.1
2.2 0.0 0.1 51.8
For no-tracking case, treatment isocenter placed in full exhale phase of target trajectory. Calculation of gamma-metric refers to same normalization value for four films (i.e., maximal dose recorded for reference measurement).
MLC tracking using electromagnetic target position monitoring d A. KRAUSS et al.
585
Fig. 7. (a,b) Gamma-test results for Films 1 and 4 displayed as dose distributions recorded on tracking film, with points failing 2%/2 mm gamma-criterion marked in red and blue for overdosage and underdosage, respectively. (c,d) Horizontal profiles of films 1 and 4 for reference, tracking, and no-tracking cases. Dose values normalized to maximal dose delivered to tumor.
regular motion traces used in our study. The tracking accuracy was primarily limited by the finite leaf width of 5 mm in the isocenter. Assuming that target motion was taken into account using appropriate margins for the three investigated scenarios, we considered the region inside the 95% isodose line of the dose distributions as the clinical target volume (CTV). Target motion resulted in a 41% decrease of the CTV. The region outside the CTV receiving #20% of the target dose was increased by 110%. Both effects were negligible in the tracking case (CTV decrease of 1%, irradiated healthy tissue increase of 5%). For a single IMRT beam applied to the phantom moving in two dimensions, the 2%/2-mm gamma-failure rate was reduced from 54.6% to 15.6% by applying the tracking technique. Again, the remaining inaccuracy could be addressed to the finite leaf width. Because several segments of the IMRT field exhibited large leaf apertures directly adjacent to a closed leaf pair, the effect of the finite leaf width on the IMRT dose distribution was by far more pronounced than
for the circular field. We will investigate whether an optimized collimator angle can help to generate field shapes that will be better suited for delivery in two-dimensional tracking mode. To investigate the clinical benefits of the tracking system, the dosimetric accuracy was assessed for a five-beam IMRT plan. The strong underdosage of the tumor edge traveling out of the treatment field without tracking applied could be compensated effectively. The 2%/2 mm gamma-failure rate decreased from 99.6% to 8.6%. In clinical practice, this underdosage would have been addressed by adding appropriate margins in the planning phase. The overdosage of the surrounding tissue traveling into the treatment beam without tracking recorded on Film 4 showed that the dose gradients intended to protect the healthy tissue from an excessive dose were severely compromised without tracking. This effect could not be completely eliminated. Nevertheless, a reduction of the 3%/3 mm gamma-failure rate from 96.7% to 51.8% and of the mean overdosage from 20.9% to 5.2% of the target dose observed for a film placed 5 mm outside the target volume showed at least considerable
586
I. J. Radiation Oncology d Biology d Physics
improvement when the tracking technique was applied. The dose profiles displayed in Fig. 7d show that the maximal dose recorded on that particular film for the reference measurement was approximately 55% of the maximal dose delivered to the tumor. This indicates that the film was placed in a region of steep dose gradients. The remaining inaccuracy for the dose delivery thus showed the sensitivity of the steep dose gradients achievable with IMRT to minor target displacements. Even for the idealized conditions of this tracking experiment, we were not able to completely restore the planned dose distribution. We are planning to perform experiments similar to those presented in the present study with phantoms moving on irregular, patient-specific motion traces. We are, therefore, currently incorporating a three-dimensional programmable motion stage into our experimental setup. Recently, the implantation of the Beacon transponders into the prostate was approved in the European Union. We aim to bring the integrated tracking system presented in the present work into clinical prostate treatment routines. Toward this goal, the thorough development of adequate quality assurance routines is essential. For lung applications, a novel transponder design is currently under development. A recent study showed 100% long-term fixation rates for these transponders implanted bronchoscopically into canine lungs (26).
Volume 79, Number 2, 2011
Another challenge to be addressed in the future is the management of more complex target motion. In the current setup, the integrated tracking system compensated in real time for three-dimensional rigid target translations according to the position information of the centroid of the three transponders. A continuous update of the individual transponder positions could, however, yield insights into possible tumor rotations or deformations. In addition, we have previously reported on synchronized target position monitoring with the Calypso System and fluoroscopic x-ray imaging (27). The combination of these techniques could allow the monitoring of target deformations and rotations and the motion of organs at risk. Future work will include strategies to realign the MLC leaves in real time to also account for this additional target information.
CONCLUSION The integrated system characterized in the present study represents a major step toward the clinical implementation of intrafractional tumor motion compensation. We could demonstrate highly accurate target tracking and showed that even complex IMRT dose distributions can be recovered to a high degree using the integrated tracking system. However, several remaining challenges need to be addressed.
REFERENCES 1. Bortfeld T, Jiang SB, Rietzel E. Effects of motion on the total dose distribution. Semin Radiat Oncol 2004;14:41–51. 2. Xing L, Thorndyke B, Schreibmann E, et al. Overview of image-guided radiation therapy. Med Dosim 2006;31: 91–112. 3. Oelfke U, Tucking T, Nill S, et al. Linac-integrated kV-cone beam CT: Technical features and first applications. Med Dosim 2006;31:62–70. 4. Sonke JJ, Rossi M, Wolthaus J, et al. Frameless stereotactic body radiotherapy for lung cancer using four-dimensional cone beam CT guidance. Int J Radiat Oncol Biol Phys 2009; 74:567–574. 5. Gierga DP, Brewer J, Sharp GC, et al. The correlation between internal and external markers for abdominal tumors: Implications for respiratory gating. Int J Radiat Oncol Biol Phys 2005;61:1551–1558. 6. Beddar AS, Kainz K, Briere TM, et al. Correlation between internal fiducial tumor motion and external marker motion for liver tumors imaged with 4D-CT. Int J Radiat Oncol Biol Phys 2007;67:630–638. 7. Shirato H, Oita M, Fujita K, et al. Feasibility of synchronization of realtime tumor-tracking radiotherapy and intensitymodulated radiotherapy from viewpoint of excessive dose from fluoroscopy. Int J Radiat Oncol Biol Phys 2004;60: 334–341. 8. Balter JM, Wright JN, Newell LJ, et al. Accuracy of a wireless localization system for radiotherapy. Int J Radiat Oncol Biol Phys 2005;61:933–937. 9. Schweikard A, Shiomi H, Adler J. Respiration tracking in radiosurgery. Med Phys 2004;31:2738–2741. 10. Seppenwoolde Y, Berbeco RI, Nishioka S, et al. Accuracy of tumor motion compensation algorithm from a robotic respira-
11. 12.
13. 14. 15. 16. 17. 18. 19. 20.
tory tracking system: A simulation study. Med Phys 2007;34: 2774–2784. D’Souza WD, McAvoy TJ. An analysis of the treatment couch and control system dynamics for respiration-induced motion compensation. Med Phys 2006;33:4701–4709. Qiu P, D’Souza WD, McAvoy TJ, et al. Inferential modelling and predictive feedback control in real-time motion compensation using the treatment couch during radiotherapy. Phys Med Biol 2007;52:5831–5854. Shirato H, Shimizu S, Kunieda T, et al. Physical aspects of a real-time tumor-tracking system for gated radiotherapy. Int J Radiat Oncol Biol Phys 2000;48:1187–1195. Shen S, Duan J, Fiveash JB, et al. Validation of target volume and position in respiratory gated CT planning and treatment. Med Phys 2003;30:3196–3205. Takayama K, Mizowaki T, Kokubo M, et al. Initial validations for pursuing irradiation using a gimbals tracking system. Radiother Oncol 2009;93:45–49. McQuaid D, Partridge M, Symonds-Tayler JR, et al. Targettracking deliveries on an Elekta linac: A feasibility study. Phys Med Biol 2009;54:3563–3578. Sawant A, Venkat R, Srivastava V, et al. Management of three dimensional intrafraction motion through real-time DMLC tracking. Med Phys 2008;35:2050–2061. Tacke MB, Nill S, Krauss A, et al. Real-time tumor tracking: Automatic compensation of target motion using the Siemens 160 MLC. Med Phys 2010;37:753–761. Low DA, Dempsey JF. Evaluation of the gamma dose distribution comparison method. Med Phys 2003;30:2455–2464. Dietrich L, Tucking T, Nill S, et al. Compensation for respiratory motion by gated radiotherapy: An experimental study. Phys Med Biol 2005;50:2405–2414.
MLC tracking using electromagnetic target position monitoring d A. KRAUSS et al.
21. Lujan AE, Larsen EW, Balter JM, et al. A method for incorporating organ motion due to breathing into 3D dose calculations. Med Phys 1999;26:715–720. 22. Sawant A, Smith RL, Venkat RB, et al. Toward submillimeter accuracy in the management of intrafraction motion: The integration of real-time internal position monitoring and multileaf collimator target tracking. Int J Radiat Oncol Biol Phys 2009; 74:575–582. 23. Hoogeman M, Prevost JB, Nuyttens J, et al. Clinical accuracy of the respiratory tumor tracking system of the CyberKnife: Assessment by analysis of log files. Int J Radiat Oncol Biol Phys 2009;74:297–303.
587
24. Murphy MJ, Dieterich S. Comparative performance of linear and nonlinear neural networks to predict irregular breathing. Phys Med Biol 2006;51:5903–5914. 25. Ruan D. Kernel density estimation-based real-time prediction for respiratory motion. Phys Med Biol 2010;55:1311–1326. 26. Mayse ML, Smith RL, Park M, et al. Development of a nonmigrating electromagnetic transponder system for lung tumor tracking. Int J Radiat Oncol Biol Phys 2008;72:S430. 27. Rau AW, Nill S, Eidens RS, et al. Synchronized tumour tracking with electromagnetic transponders and kV x-ray imaging: Evaluation based on a thorax phantom. Phys Med Biol 2008; 53:3789–3805.