- Email: [email protected]
Ultrasound tracking for intra-fractional motion compensation in radiation therapy
Physica Medica xxx (2014) 1e5
Contents lists available at ScienceDirect
Physica Medica journal homepage: http://www.physicamedica.com
Original paper
Ultrasound tracking for intra-fractional motion compensation in radiation therapy J. Schwaab a, *, M. Prall b, C. Sarti a, R. Kaderka b, C. Bert b, C. Kurz c, K. Parodi c, d, M. Günther a, e, J. Jenne a, e a
Mediri GmbH, Heidelberg, Germany GSI Helmholtzzentrum für Schwerionenforschung GmbH, Darmstadt, Germany Heidelberg Ion-Beam Therapy Center (HIT) and Department of Radiation Oncology, Heidelberg University Hospital, Germany d Ludwig Maximilian University, Munich, Germany e Fraunhofer MEVIS, Bremen, Germany b c
a r t i c l e i n f o
a b s t r a c t
Article history: Received 15 January 2014 Received in revised form 6 March 2014 Accepted 9 March 2014 Available online xxx
Modern techniques as ion beam therapy or 4D imaging require precise target position information. However, target motion particularly in the abdomen due to respiration or patient movement is still a challenge and demands methods that detect and compensate this motion. Ultrasound represents a noninvasive, dose-free and model-independent alternative to fluoroscopy, respiration belt or optical tracking of the patient surface. Thus, ultrasound based motion tracking was integrated into irradiation with actively scanned heavy ions. In a first in vitro experiment, the ultrasound tracking system was used to compensate diverse sinusoidal target motions in two dimensions. A time delay of w200 ms between target motion and reported position data was compensated by a prediction algorithm (artificial neural network). The irradiated films proved feasibility of the proposed method. Furthermore, a practicable and reliable calibration workflow was developed to enable the transformation of ultrasound tracking data to the coordinates of the treatment delivery or imaging system e even if the ultrasound probe moves due to respiration. A first proof of principle experiment was performed during time-resolved positron emission tomography (4DPET) to test the calibration workflow and to show the accuracy of an ultrasound based motion tracking in vitro. The results showed that optical ultrasound tracking can reach acceptable accuracies and encourage further research. Ó 2014 Associazione Italiana di Fisica Medica. Published by Elsevier Ltd. All rights reserved.
Keywords: Ultrasound Motion correction Particle therapy
Introduction Modern techniques in radiation therapy such as ion beam therapy or intensity modulated radiation therapy (IMRT) can achieve delivery accuracies at the millimeter scale. However, target motion particularly in the abdomen due to patient movement or respiration imposes inaccuracies that cannot be neglected [1]. In addition, these difficulties arise not only in radiation therapy but also in diagnostic imaging. Therefore, methods that detect and compensate this motion are in strong demand [2]. Many approaches have been proposed already to avoid substantial dose errors and distorted images. Optical systems [3] or breathing belts e as used in the Respiratory Gating System AZ-733V (ANZAI
* Corresponding author. Tel.: þ49 62217256975. E-mail address: [email protected] (J. Schwaab).
Medical Co., Ltd., Tokyo, Japan) e generally only return 1D tracking information of the abdominal surface. Fluoroscopy yields the desired information on inner structures but implies an additional radiation burden for the patient [4]. The Calypso System (Calypso Medical Technology, Seattle, WA) which is used in prostate radiation therapy uses implanted RF-transponders for continuous motion tracking of a tumor [5]. However, this is highly invasive as fiducials (small beacons) have to be implanted accurately near the tumor. Here diagnostic ultrasound (sonography) represents a noninvasive, dose-free, model-independent, real time capable and cheap alternative. First experiments to integrate ultrasound tracking to radio surgery using the CyberKnife (Accuray Inc., USA) have already been performed successfully. Blanck et al. [6] replaced the standard optical tracking system by a GE Vivid 7 (General Electronics, USA) ultrasound device to follow the motion of fiducials in a water bath. The fiducials were coupled rigidly to a simple
http://dx.doi.org/10.1016/j.ejmp.2014.03.003 1120-1797/Ó 2014 Associazione Italiana di Fisica Medica. Published by Elsevier Ltd. All rights reserved.
Please cite this article in press as: Schwaab J, et al., Ultrasound tracking for intra-fractional motion compensation in radiation therapy, Physica Medica (2014), http://dx.doi.org/10.1016/j.ejmp.2014.03.003
2
J. Schwaab et al. / Physica Medica xxx (2014) 1e5
beam target and moved by a robotic arm. Schlosser et al. [7] combined telerobotic diagnostic ultrasound with projection X-ray imaging in image guided radiation therapy (IGRT). The two modalities complemented each other during IGRT delivery on a tissue mimicking phantom with gold fiducial markers. While these two approaches rely on the tracking of fiducial markers that might need to be implanted to the patient, the final goal of this project is to develop an ultrasound based motion tracking for real time motion correction notably in ion beam therapy that works without markers. It is favorable to avoid the use of fiducial markers as they require an implantation and might influence the dose distribution during therapy [8]. In this work, two first steps towards this goal were performed in vitro: First, ultrasound tracking was integrated into treatment delivery with actively scanned heavy ions [9]. Thus, this study represents a new technical combination of two modalities. Secondly, a workflow was developed to enable the transformation of ultrasound tracking data to the coordinates of the treatment delivery system e even if the ultrasound probe was moved itself during acquisition. This may be the case if the probe is attached directly to the patient’s skin instead of being hold by a static holder. Material and methods The ultrasound tracking software used, Sonoplan II, is developed by mediri GmbH, Heidelberg, Germany, and runs on a DiPhAS (Digital Phased Array System, Fraunhofer IBMT, St. Ingbert, Germany) ultrasound device. The DiPhAS comprises the beam former and a PC such that the necessary raw data can be accessed easily and processed directly on the machine. The ultrasound probe that was used during these experiments consists of two orthogonally oriented phased arrays with 64 elements, respectively. This allows quasi-simultaneous imaging in two perpendicular image planes. The tracking algorithm uses active contours and conditional density propagation as described in detail by Zhang et al. [10]. Based on the brightness values of a manually segmented structure, the algorithm yields up to five features of the moving target in real-time (5 ms per frame): translation and scaling in x- and y-direction, respectively, as well as rotation within the xey-plane. Taking all image processing and tracking calculations together, the software needs about 40 ms to calculate tracking data out of the incoming raw data of one frame. As the two image planes are generated alternately, it takes about 80 ms to get the complete tracking information of both the frames. In our experiment, additional image processing and data transfer operations lead to a sampling rate of 150 ms for the two frames.
Figure 1. Experimental setup for ultrasound tracking in ion beam therapy: Both, the radiosensitive film and the rubber ball were moved by the robotic arm in two directions perpendicular to the ion beam. The motion was detected by the ultrasound tracking system and transferred to the treatment delivery system.
was several minutes. This brings out again the need for a motion correction that works reliably over a time span of several minutes. The films were attached to the same robotic arm as the rubber ball. Thus, the motions of the tracking target and the beam target were coupled directly. The displacement of the rubber ball was continuously measured by the US tracking system and sent to the therapy control system (TCS). Hence, the study relies on the already established TCS which is indeed capable to track (steer) the ion beam according to incoming target position information in realtime. However, to our knowledge, it is the first time that ultrasound tracking information was used to adapt the beam position. Image processing, tracking calculations and data communication introduced a delay of w200 ms leading to a position error of several millimeters. An artificial neural network (ANN) was implemented in the TCS to predict motion from US measurements and thus to compensate for the delay. Note, that the tracking information for both ultrasound image planes was generated, but as the target motion was two dimensional, we only used the tracking data of the plane perpendicular to the beam. The irradiated films were developed and digitized with a resolution of 150 pix/inch (1 pix ¼ 0.17 mm) and 16 bit depth using a medical film scanner (Vidar DosimetryPro Advantage). Tracking in absolute coordinates
Integration into active ion beam delivery As this experiment should serve as a first proof of principle for the use of ultrasound tracking information in active ion beam delivery, a straightforward setup was chosen. A rubber ball target inside a water bath was moved by a robotic arm in two dimensions orthogonal to the ion beam at GSI Helmholtzzentrum für Schwerionenforschung GmbH, Darmstadt, Germany (see Fig. 1). The robotic arm generated various sinusoidal trajectories having a period of 3s and a peak-to-peak amplitude of 20 mm. Squares of 3 3 cm2 should be homogeneously irradiated by a calcium ion beam with an energy of 200 MeV/u and a beam width (FWHM) of 6mm on radiosensitive films using the beam tracking (steering) technique. Each field was irradiated by one single scan, the distance between the single beam spots was 2 mm in both directions. Thus, each field was created by 15 15 ¼ 225 beam spots. The mean irradiation time for one beam spot was around 1 ms, nevertheless, caused by the ion beam delivery system, the total irradiation time for one field
In order to perform ultrasound tracking, it is necessary that the probe stays in contact with the patient. This can be achieved either by a static holder in combination with a gel pad that compensates breathing motion or by a probe that is attached directly to the skin. In the latter case, the probe position has to be registered to allow tracking in absolute coordinates. For this part of the study, the ultrasound probe was equipped with an optical marker in order to be detected by an optical tracking system, the Passive Polaris Spectra measurement system (Northern Digital Inc., Waterloo, Ontario). After calibration of both, the ultrasound system and the optical tracking system, the setup was tested in a PET/CT (Positron Emission Tomography/Computed Tomography) scanner at the Heidelberg Ion-Beam Therapy Center (HIT) to show the feasibility of the proposed workflow. The target was a combination of a rubber ball in a water bath and a radioactive point source. The target was moved by a QUASAR respiratory phantom (Modus Medical Devices Inc., London, Canada) with a peak-to-peak amplitude of 40 mm and
Please cite this article in press as: Schwaab J, et al., Ultrasound tracking for intra-fractional motion compensation in radiation therapy, Physica Medica (2014), http://dx.doi.org/10.1016/j.ejmp.2014.03.003
J. Schwaab et al. / Physica Medica xxx (2014) 1e5
3
a period of 3 s along the symmetry axis of the scanner. At the same time, the probe was moved perpendicularly by an ANZAI respiratory phantom with a peak-to-peak amplitude of 20 mm and a period of 4 s. The target was designed such that it could be detected by the moving ultrasound probe as well as by the PET/CT scanner for comparison of the results (see Fig. 2). Results The following results confirm that both the presented steps towards an integration of ultrasound tracking for ion beam therapy proved to be practicable. Integration into active ion beam delivery In this part of the study, a moving radiosensitive film was irradiated by an actively scanned calcium ion beam. The beam tracking
Figure 3. Irradiated film. The interplay effect is clearly visible when the target is moving and no prediction is done (similar to the case when no tracking is done at all). However, the fields are nearly homogeneous when artificial neural network prediction is used.
(steering) was based on the relative coordinates that were extracted from the ultrasound tracking information. An artificial neural network was implemented and e after a training phase of 30 s e predicted the target motion in order to compensate for the time delay that arose in data processing and transfer. Five future positions were calculated from the ten latest ultrasound tracking positions. A static irradiation without target motion was performed as reference and produced the expected homogeneous square field (see Fig. 3, top left). Then, the target was moved on three different 2D sinusoidal trajectories (sin, sin2, sin4) orthogonal to the ion beam direction and irradiated once with motion correction but without prediction and once with both of them, respectively. The irradiation without prediction yielded the pure interplay effect similar to the case without any motion correction. Thus, steering the beam according to the ultrasound tracking information but without the neural network prediction produced distorted, irregular irradiation fields, caused by the considerable delay in data transfer (see Fig. 3, top). However, the combination of ultrasound tracking with neural network prediction yielded satisfying results as shown in Fig. 3, bottom, where the homogeneity and spreading of the irradiation patterns are clearly enhanced.
Tracking in absolute coordinates
Figure 2. Experimental setup for ultrasound tracking in treatment room coordinates. Both, the target (rubber ball and radioactive source) and the probe are moved by respiratory phantoms. The probe motion was tracked optically. The target motion was sinusoidal with 40 mm amplitude and 3 s period. The probe was moved orthogonally to the target with 20 mm amplitude and 4 s period.
After thorough calibration of the optical and the ultrasound tracking system, the trajectory of a moving rubber ball was measured by the proposed combined tracking system: The ultrasound tracking system measured the position of the rubber ball while the moving ultrasound probe was detected by the optical tracking system. This way, the target motion could be reconstructed in absolute coordinates. Figure 4 shows the tracks of the rubber ball in both the coordinate systems of the ultrasound device and of the treatment room (target coordinate system). The temporally dephased superposition of target and probe motion is clearly visible to the expected extent in the left plot (ultrasound coordinates): Along the xUS-axis, one can see the probe motion with an amplitude of 20 mm (and a period of 4 s) and along the yUS-axis, the target motion is present with an amplitude of 40 mm (and a period of 3 s). As shown in the right plot, the influence of the probe motion was enormously diminished in the transformed tracking data which, thus, represents the trajectory of the target in treatment room coordinates. While the target motion with its original amplitude of 40 mm is clearly visible along the zW-axis, the probe motion is reduced by 80%. Only a slight uncertainty of about 4 mm along the xW-axis remained which, however, is within the overall uncertainty of the combined tracking system. PET/CT reconstruction images confirmed the detected target motion.
Please cite this article in press as: Schwaab J, et al., Ultrasound tracking for intra-fractional motion compensation in radiation therapy, Physica Medica (2014), http://dx.doi.org/10.1016/j.ejmp.2014.03.003
4
J. Schwaab et al. / Physica Medica xxx (2014) 1e5
ion beam range shifting, the motion correction in the presented work was only in 2D although the ultrasound tracking yielded 3D position information. The same accounts for Blanck et al. who used the integrated Synchrony module of the CyberKnife which can only handle 2D motion data. The target motion was sinusoidal in two dimensions with peak-to-peak amplitudes of 20 mm and a period of 3 s and, thus, represents an upper limit for realistic respiratory patient movement. Nevertheless, the motion correction yielded satisfying results. Blanck et al. used a superposition of 15 mm respiratory motion and 5 mm cardiac pulses. Their tracking frame rate was 17.7 Hz (w55 ms per frame), while the Sonoplan II system ran at a frame rate of 6.7 Hz (w150 ms per frame) when the experiment was performed. However, this is still acceptable considering the typical velocities of patient motion. Several contributions to the processing time per frame have been investigated. The finite speed of sound in water (v ¼ 1540 m/s) makes the ultrasound waves of both image planes (2 64 lines) travel for 25 ms at a scanning depth of 15 cm. Image processing and target tracking takes about 80 ms for both image planes. In previous measurements, the duration of data transfer via UDP network was found to be 16 7 ms [13]. It is probable that non-optimized data handling in the ultrasound imaging and tracking software is the reason for the residual 70 ms. This delay has to be further investigated and may be reduced by improving the communication between beam former and image processing unit. Additionally, the data transfer between DiPhAS and TCS via UDP network could be sub-optimal. Tracking in absolute coordinates
Figure 4. Trajectories of the rubber ball in ultrasound coordinates (left) and transformed to treatment room coordinates (right). While both, the probe and target motions are obvious in the left plot, the probe motion is nearly eliminated in the right plot.
Discussion Integration into active ion beam delivery The ultrasound tracking system implemented in this study proved to be reliable and feasible in combination with ion beam delivery. The presented experiment was comparable to the setups of Blanck et al. [6] and Schlosser et al. [7]. While the latter one used ultrasound tracking only as a complement to X-ray positioning which required implanted markers, our approach is a variable, stand-alone ultrasound device that will be able to yield absolute target coordinate information. The presented ultrasound tracking system works without implanted markers which could influence the dose distribution. Furthermore, the system has already been tested in combination with diagnostic imaging devices: In a phantom study, motion correction during 4D PET (positron emission tomography) imaging for treatment verification [11] was performed. Moreover an MRI (magnetic resonance imaging) compatible version of the system has been developed for motion compensation in 4D MRI [12]. Due to the complexity of real-time
The presented data prove that the probe motion can be detected and compensated reliably by the used calibration method. The target motion was reproduced in the correct direction with the expected amplitude and the probe movement was mitigated by 80% due to the optical tracking. Considering that the chosen motion amplitudes were rather representing an upper limit in respiratory motion, this result is promising. Although the overall accuracy of the tracking system is only slightly below 5 mm, it is still acceptable taking into account that the tracking information is 2D and is acquired in situ. To our knowledge, established non-ionizing portable systems mostly yield 1D information and track the patient surface or other surrogates. Conclusion In this work, it could be shown in vitro that ultrasound tracking is a practicable motion mitigation method for ion beam therapy. It can yield nearly real-time position information at high frame rate of moving targets (instead of surrogates like chest wall motion). Due to the calibrated optical tracking of the probe that was implemented in this work, target motion can be recorded correctly even if the probe is moved itself. In future work, the time delay between ultrasound and beam delivery has to be investigated thoroughly and the quasi 3D capability of the ultrasound tracking will be exploited by irradiating 3D fields. The latter one requires adding beam range adaption (e.g. using a mechanical range shifter [14]) to the beam deflection. Furthermore, the setup for tracking in absolute coordinates has to be tested in vivo. Acknowledgments Parts of this work were supported by the EU FP7 Project ENVISION (European NoVel Imaging Systems for ION therapy) under grant agreement no. 241851, and by the Bundesministerium für Bildung und Forschung BIO-DISC 5 Project no. 0315726A.
Please cite this article in press as: Schwaab J, et al., Ultrasound tracking for intra-fractional motion compensation in radiation therapy, Physica Medica (2014), http://dx.doi.org/10.1016/j.ejmp.2014.03.003
J. Schwaab et al. / Physica Medica xxx (2014) 1e5
References [1] Shirato H, Shimizu S, Kitamura K, Onimaru R. Organ motion in image-guided radiotherapy: lessons from real-time tumor-tracking radiotherapy. Int J Clin Oncol 2007;12:8. [2] Bert C, Durante M. Motion in radiotherapy: particle therapy. Phys Med Biol 2011;56:R113. [3] Schaerer J, Fassi A, Riboldi M, Ceveri P, Baroni G, Sarrut D. Multi-dimensional respiratory motion tracking from markerless optical surface imaging based on deformable mesh registration. Phys Med Biol 2012;57:357. [4] Berbeco RI, Jiang SB, Sharp GC, Chen GT, Mostafavi H, Shirato H. Integrated radiotherapy imaging system (IRIS): design considerations of tumour tracking with linac gantry-mounted diagnostic x-ray systems with flat-panel detectors. Phys Med Biol 2004;49:243. [5] Kupelian P, Willoughby T, Mahadevan A, Djemil T, Weinstein G, Jani S, et al. Multi-institutional clinical experience with the Calypso system in localization and continuous, real-time monitoring of the prostate gland during external radiotherapy. Int J Radiat Oncol Biol Phys 2007;67(4):1088e98. [6] Blanck O, Jauer P, Ernst F, Bruder R, Schweikard A. Pilot-Phantomstudie zur ultraschallgeführten, robotergestützten Radiochirurgie. In: 44. Jahrestagung der Deutschen Gesellschaft für Medizinische Physik (DGMP); 2013. p. 122. [7] Schlosser J, Salisbury K, Xing L, Hritov D. Hybrid X-ray/ultrasound imaging approach for patient positioning and real-time tracking in IGRT. Med Phys 2011;38:3389.
5
[8] Jäckel O, Reiss P. The influence of metal artifacts on the range of ion beams. Phys Med Biol 2007;52:635e44. [9] Saito N, Bert C, Chaudhri N, Gemmel A, Schardt D, Durante M, et al. Speed and accuracy of a beam tracking system for treatment of moving targets with scanned ion beams. Phys Med Biol 2009;54:4849. [10] Zhang X, Günther M, Bongers A. Real-time organ motion tracking in ultrasound imaging using active contours and conditional density propagation. Medical Imaging and Augmented Reality Lecture Notes Comput Sci 2010;6326:286. [11] Kurz C, Bauer J, Bert C, Bongers A, Jenne J, Richter D, et al. First steps towards 4D offline PET-based treatment verification at the Heidelberg Ion Beam Therapy Center. Radiother Oncol 2012;102:55. [12] Tretbar S, Hewener H, Speicher D, Bartscherer T, Bongers A, Jenne J, et al. MR-compatible ultrasound research platform for motion tracking to reduce motion induced artifacts in MR imaging. In: 2013 IEEE International Ultrasonics Symposium (IUS); 2013. p. 553. [13] Fattori G, Saito N, Pella A, Kaderka R, Seregni M, Constantinescu A, et al. TU-ABRA-08: integration of optical tracking for organ motion compensation in scanned ion-beam therapy. Med Phys 2012;39:3889. [14] Chaudhri N, Saito N, Bert C, Franczak B, Steidl B, Durante M, et al. Ion-optical studies for a range adaptation method in ion beam therapy using a static wedge degrader combined with magnetic beam deflection. Phys Med Biol 2010;55:3499e513.
Please cite this article in press as: Schwaab J, et al., Ultrasound tracking for intra-fractional motion compensation in radiation therapy, Physica Medica (2014), http://dx.doi.org/10.1016/j.ejmp.2014.03.003
Contents lists available at ScienceDirect
Physica Medica journal homepage: http://www.physicamedica.com
Original paper
Ultrasound tracking for intra-fractional motion compensation in radiation therapy J. Schwaab a, *, M. Prall b, C. Sarti a, R. Kaderka b, C. Bert b, C. Kurz c, K. Parodi c, d, M. Günther a, e, J. Jenne a, e a
Mediri GmbH, Heidelberg, Germany GSI Helmholtzzentrum für Schwerionenforschung GmbH, Darmstadt, Germany Heidelberg Ion-Beam Therapy Center (HIT) and Department of Radiation Oncology, Heidelberg University Hospital, Germany d Ludwig Maximilian University, Munich, Germany e Fraunhofer MEVIS, Bremen, Germany b c
a r t i c l e i n f o
a b s t r a c t
Article history: Received 15 January 2014 Received in revised form 6 March 2014 Accepted 9 March 2014 Available online xxx
Modern techniques as ion beam therapy or 4D imaging require precise target position information. However, target motion particularly in the abdomen due to respiration or patient movement is still a challenge and demands methods that detect and compensate this motion. Ultrasound represents a noninvasive, dose-free and model-independent alternative to fluoroscopy, respiration belt or optical tracking of the patient surface. Thus, ultrasound based motion tracking was integrated into irradiation with actively scanned heavy ions. In a first in vitro experiment, the ultrasound tracking system was used to compensate diverse sinusoidal target motions in two dimensions. A time delay of w200 ms between target motion and reported position data was compensated by a prediction algorithm (artificial neural network). The irradiated films proved feasibility of the proposed method. Furthermore, a practicable and reliable calibration workflow was developed to enable the transformation of ultrasound tracking data to the coordinates of the treatment delivery or imaging system e even if the ultrasound probe moves due to respiration. A first proof of principle experiment was performed during time-resolved positron emission tomography (4DPET) to test the calibration workflow and to show the accuracy of an ultrasound based motion tracking in vitro. The results showed that optical ultrasound tracking can reach acceptable accuracies and encourage further research. Ó 2014 Associazione Italiana di Fisica Medica. Published by Elsevier Ltd. All rights reserved.
Keywords: Ultrasound Motion correction Particle therapy
Introduction Modern techniques in radiation therapy such as ion beam therapy or intensity modulated radiation therapy (IMRT) can achieve delivery accuracies at the millimeter scale. However, target motion particularly in the abdomen due to patient movement or respiration imposes inaccuracies that cannot be neglected [1]. In addition, these difficulties arise not only in radiation therapy but also in diagnostic imaging. Therefore, methods that detect and compensate this motion are in strong demand [2]. Many approaches have been proposed already to avoid substantial dose errors and distorted images. Optical systems [3] or breathing belts e as used in the Respiratory Gating System AZ-733V (ANZAI
* Corresponding author. Tel.: þ49 62217256975. E-mail address: [email protected] (J. Schwaab).
Medical Co., Ltd., Tokyo, Japan) e generally only return 1D tracking information of the abdominal surface. Fluoroscopy yields the desired information on inner structures but implies an additional radiation burden for the patient [4]. The Calypso System (Calypso Medical Technology, Seattle, WA) which is used in prostate radiation therapy uses implanted RF-transponders for continuous motion tracking of a tumor [5]. However, this is highly invasive as fiducials (small beacons) have to be implanted accurately near the tumor. Here diagnostic ultrasound (sonography) represents a noninvasive, dose-free, model-independent, real time capable and cheap alternative. First experiments to integrate ultrasound tracking to radio surgery using the CyberKnife (Accuray Inc., USA) have already been performed successfully. Blanck et al. [6] replaced the standard optical tracking system by a GE Vivid 7 (General Electronics, USA) ultrasound device to follow the motion of fiducials in a water bath. The fiducials were coupled rigidly to a simple
http://dx.doi.org/10.1016/j.ejmp.2014.03.003 1120-1797/Ó 2014 Associazione Italiana di Fisica Medica. Published by Elsevier Ltd. All rights reserved.
Please cite this article in press as: Schwaab J, et al., Ultrasound tracking for intra-fractional motion compensation in radiation therapy, Physica Medica (2014), http://dx.doi.org/10.1016/j.ejmp.2014.03.003
2
J. Schwaab et al. / Physica Medica xxx (2014) 1e5
beam target and moved by a robotic arm. Schlosser et al. [7] combined telerobotic diagnostic ultrasound with projection X-ray imaging in image guided radiation therapy (IGRT). The two modalities complemented each other during IGRT delivery on a tissue mimicking phantom with gold fiducial markers. While these two approaches rely on the tracking of fiducial markers that might need to be implanted to the patient, the final goal of this project is to develop an ultrasound based motion tracking for real time motion correction notably in ion beam therapy that works without markers. It is favorable to avoid the use of fiducial markers as they require an implantation and might influence the dose distribution during therapy [8]. In this work, two first steps towards this goal were performed in vitro: First, ultrasound tracking was integrated into treatment delivery with actively scanned heavy ions [9]. Thus, this study represents a new technical combination of two modalities. Secondly, a workflow was developed to enable the transformation of ultrasound tracking data to the coordinates of the treatment delivery system e even if the ultrasound probe was moved itself during acquisition. This may be the case if the probe is attached directly to the patient’s skin instead of being hold by a static holder. Material and methods The ultrasound tracking software used, Sonoplan II, is developed by mediri GmbH, Heidelberg, Germany, and runs on a DiPhAS (Digital Phased Array System, Fraunhofer IBMT, St. Ingbert, Germany) ultrasound device. The DiPhAS comprises the beam former and a PC such that the necessary raw data can be accessed easily and processed directly on the machine. The ultrasound probe that was used during these experiments consists of two orthogonally oriented phased arrays with 64 elements, respectively. This allows quasi-simultaneous imaging in two perpendicular image planes. The tracking algorithm uses active contours and conditional density propagation as described in detail by Zhang et al. [10]. Based on the brightness values of a manually segmented structure, the algorithm yields up to five features of the moving target in real-time (5 ms per frame): translation and scaling in x- and y-direction, respectively, as well as rotation within the xey-plane. Taking all image processing and tracking calculations together, the software needs about 40 ms to calculate tracking data out of the incoming raw data of one frame. As the two image planes are generated alternately, it takes about 80 ms to get the complete tracking information of both the frames. In our experiment, additional image processing and data transfer operations lead to a sampling rate of 150 ms for the two frames.
Figure 1. Experimental setup for ultrasound tracking in ion beam therapy: Both, the radiosensitive film and the rubber ball were moved by the robotic arm in two directions perpendicular to the ion beam. The motion was detected by the ultrasound tracking system and transferred to the treatment delivery system.
was several minutes. This brings out again the need for a motion correction that works reliably over a time span of several minutes. The films were attached to the same robotic arm as the rubber ball. Thus, the motions of the tracking target and the beam target were coupled directly. The displacement of the rubber ball was continuously measured by the US tracking system and sent to the therapy control system (TCS). Hence, the study relies on the already established TCS which is indeed capable to track (steer) the ion beam according to incoming target position information in realtime. However, to our knowledge, it is the first time that ultrasound tracking information was used to adapt the beam position. Image processing, tracking calculations and data communication introduced a delay of w200 ms leading to a position error of several millimeters. An artificial neural network (ANN) was implemented in the TCS to predict motion from US measurements and thus to compensate for the delay. Note, that the tracking information for both ultrasound image planes was generated, but as the target motion was two dimensional, we only used the tracking data of the plane perpendicular to the beam. The irradiated films were developed and digitized with a resolution of 150 pix/inch (1 pix ¼ 0.17 mm) and 16 bit depth using a medical film scanner (Vidar DosimetryPro Advantage). Tracking in absolute coordinates
Integration into active ion beam delivery As this experiment should serve as a first proof of principle for the use of ultrasound tracking information in active ion beam delivery, a straightforward setup was chosen. A rubber ball target inside a water bath was moved by a robotic arm in two dimensions orthogonal to the ion beam at GSI Helmholtzzentrum für Schwerionenforschung GmbH, Darmstadt, Germany (see Fig. 1). The robotic arm generated various sinusoidal trajectories having a period of 3s and a peak-to-peak amplitude of 20 mm. Squares of 3 3 cm2 should be homogeneously irradiated by a calcium ion beam with an energy of 200 MeV/u and a beam width (FWHM) of 6mm on radiosensitive films using the beam tracking (steering) technique. Each field was irradiated by one single scan, the distance between the single beam spots was 2 mm in both directions. Thus, each field was created by 15 15 ¼ 225 beam spots. The mean irradiation time for one beam spot was around 1 ms, nevertheless, caused by the ion beam delivery system, the total irradiation time for one field
In order to perform ultrasound tracking, it is necessary that the probe stays in contact with the patient. This can be achieved either by a static holder in combination with a gel pad that compensates breathing motion or by a probe that is attached directly to the skin. In the latter case, the probe position has to be registered to allow tracking in absolute coordinates. For this part of the study, the ultrasound probe was equipped with an optical marker in order to be detected by an optical tracking system, the Passive Polaris Spectra measurement system (Northern Digital Inc., Waterloo, Ontario). After calibration of both, the ultrasound system and the optical tracking system, the setup was tested in a PET/CT (Positron Emission Tomography/Computed Tomography) scanner at the Heidelberg Ion-Beam Therapy Center (HIT) to show the feasibility of the proposed workflow. The target was a combination of a rubber ball in a water bath and a radioactive point source. The target was moved by a QUASAR respiratory phantom (Modus Medical Devices Inc., London, Canada) with a peak-to-peak amplitude of 40 mm and
Please cite this article in press as: Schwaab J, et al., Ultrasound tracking for intra-fractional motion compensation in radiation therapy, Physica Medica (2014), http://dx.doi.org/10.1016/j.ejmp.2014.03.003
J. Schwaab et al. / Physica Medica xxx (2014) 1e5
3
a period of 3 s along the symmetry axis of the scanner. At the same time, the probe was moved perpendicularly by an ANZAI respiratory phantom with a peak-to-peak amplitude of 20 mm and a period of 4 s. The target was designed such that it could be detected by the moving ultrasound probe as well as by the PET/CT scanner for comparison of the results (see Fig. 2). Results The following results confirm that both the presented steps towards an integration of ultrasound tracking for ion beam therapy proved to be practicable. Integration into active ion beam delivery In this part of the study, a moving radiosensitive film was irradiated by an actively scanned calcium ion beam. The beam tracking
Figure 3. Irradiated film. The interplay effect is clearly visible when the target is moving and no prediction is done (similar to the case when no tracking is done at all). However, the fields are nearly homogeneous when artificial neural network prediction is used.
(steering) was based on the relative coordinates that were extracted from the ultrasound tracking information. An artificial neural network was implemented and e after a training phase of 30 s e predicted the target motion in order to compensate for the time delay that arose in data processing and transfer. Five future positions were calculated from the ten latest ultrasound tracking positions. A static irradiation without target motion was performed as reference and produced the expected homogeneous square field (see Fig. 3, top left). Then, the target was moved on three different 2D sinusoidal trajectories (sin, sin2, sin4) orthogonal to the ion beam direction and irradiated once with motion correction but without prediction and once with both of them, respectively. The irradiation without prediction yielded the pure interplay effect similar to the case without any motion correction. Thus, steering the beam according to the ultrasound tracking information but without the neural network prediction produced distorted, irregular irradiation fields, caused by the considerable delay in data transfer (see Fig. 3, top). However, the combination of ultrasound tracking with neural network prediction yielded satisfying results as shown in Fig. 3, bottom, where the homogeneity and spreading of the irradiation patterns are clearly enhanced.
Tracking in absolute coordinates
Figure 2. Experimental setup for ultrasound tracking in treatment room coordinates. Both, the target (rubber ball and radioactive source) and the probe are moved by respiratory phantoms. The probe motion was tracked optically. The target motion was sinusoidal with 40 mm amplitude and 3 s period. The probe was moved orthogonally to the target with 20 mm amplitude and 4 s period.
After thorough calibration of the optical and the ultrasound tracking system, the trajectory of a moving rubber ball was measured by the proposed combined tracking system: The ultrasound tracking system measured the position of the rubber ball while the moving ultrasound probe was detected by the optical tracking system. This way, the target motion could be reconstructed in absolute coordinates. Figure 4 shows the tracks of the rubber ball in both the coordinate systems of the ultrasound device and of the treatment room (target coordinate system). The temporally dephased superposition of target and probe motion is clearly visible to the expected extent in the left plot (ultrasound coordinates): Along the xUS-axis, one can see the probe motion with an amplitude of 20 mm (and a period of 4 s) and along the yUS-axis, the target motion is present with an amplitude of 40 mm (and a period of 3 s). As shown in the right plot, the influence of the probe motion was enormously diminished in the transformed tracking data which, thus, represents the trajectory of the target in treatment room coordinates. While the target motion with its original amplitude of 40 mm is clearly visible along the zW-axis, the probe motion is reduced by 80%. Only a slight uncertainty of about 4 mm along the xW-axis remained which, however, is within the overall uncertainty of the combined tracking system. PET/CT reconstruction images confirmed the detected target motion.
Please cite this article in press as: Schwaab J, et al., Ultrasound tracking for intra-fractional motion compensation in radiation therapy, Physica Medica (2014), http://dx.doi.org/10.1016/j.ejmp.2014.03.003
4
J. Schwaab et al. / Physica Medica xxx (2014) 1e5
ion beam range shifting, the motion correction in the presented work was only in 2D although the ultrasound tracking yielded 3D position information. The same accounts for Blanck et al. who used the integrated Synchrony module of the CyberKnife which can only handle 2D motion data. The target motion was sinusoidal in two dimensions with peak-to-peak amplitudes of 20 mm and a period of 3 s and, thus, represents an upper limit for realistic respiratory patient movement. Nevertheless, the motion correction yielded satisfying results. Blanck et al. used a superposition of 15 mm respiratory motion and 5 mm cardiac pulses. Their tracking frame rate was 17.7 Hz (w55 ms per frame), while the Sonoplan II system ran at a frame rate of 6.7 Hz (w150 ms per frame) when the experiment was performed. However, this is still acceptable considering the typical velocities of patient motion. Several contributions to the processing time per frame have been investigated. The finite speed of sound in water (v ¼ 1540 m/s) makes the ultrasound waves of both image planes (2 64 lines) travel for 25 ms at a scanning depth of 15 cm. Image processing and target tracking takes about 80 ms for both image planes. In previous measurements, the duration of data transfer via UDP network was found to be 16 7 ms [13]. It is probable that non-optimized data handling in the ultrasound imaging and tracking software is the reason for the residual 70 ms. This delay has to be further investigated and may be reduced by improving the communication between beam former and image processing unit. Additionally, the data transfer between DiPhAS and TCS via UDP network could be sub-optimal. Tracking in absolute coordinates
Figure 4. Trajectories of the rubber ball in ultrasound coordinates (left) and transformed to treatment room coordinates (right). While both, the probe and target motions are obvious in the left plot, the probe motion is nearly eliminated in the right plot.
Discussion Integration into active ion beam delivery The ultrasound tracking system implemented in this study proved to be reliable and feasible in combination with ion beam delivery. The presented experiment was comparable to the setups of Blanck et al. [6] and Schlosser et al. [7]. While the latter one used ultrasound tracking only as a complement to X-ray positioning which required implanted markers, our approach is a variable, stand-alone ultrasound device that will be able to yield absolute target coordinate information. The presented ultrasound tracking system works without implanted markers which could influence the dose distribution. Furthermore, the system has already been tested in combination with diagnostic imaging devices: In a phantom study, motion correction during 4D PET (positron emission tomography) imaging for treatment verification [11] was performed. Moreover an MRI (magnetic resonance imaging) compatible version of the system has been developed for motion compensation in 4D MRI [12]. Due to the complexity of real-time
The presented data prove that the probe motion can be detected and compensated reliably by the used calibration method. The target motion was reproduced in the correct direction with the expected amplitude and the probe movement was mitigated by 80% due to the optical tracking. Considering that the chosen motion amplitudes were rather representing an upper limit in respiratory motion, this result is promising. Although the overall accuracy of the tracking system is only slightly below 5 mm, it is still acceptable taking into account that the tracking information is 2D and is acquired in situ. To our knowledge, established non-ionizing portable systems mostly yield 1D information and track the patient surface or other surrogates. Conclusion In this work, it could be shown in vitro that ultrasound tracking is a practicable motion mitigation method for ion beam therapy. It can yield nearly real-time position information at high frame rate of moving targets (instead of surrogates like chest wall motion). Due to the calibrated optical tracking of the probe that was implemented in this work, target motion can be recorded correctly even if the probe is moved itself. In future work, the time delay between ultrasound and beam delivery has to be investigated thoroughly and the quasi 3D capability of the ultrasound tracking will be exploited by irradiating 3D fields. The latter one requires adding beam range adaption (e.g. using a mechanical range shifter [14]) to the beam deflection. Furthermore, the setup for tracking in absolute coordinates has to be tested in vivo. Acknowledgments Parts of this work were supported by the EU FP7 Project ENVISION (European NoVel Imaging Systems for ION therapy) under grant agreement no. 241851, and by the Bundesministerium für Bildung und Forschung BIO-DISC 5 Project no. 0315726A.
Please cite this article in press as: Schwaab J, et al., Ultrasound tracking for intra-fractional motion compensation in radiation therapy, Physica Medica (2014), http://dx.doi.org/10.1016/j.ejmp.2014.03.003
J. Schwaab et al. / Physica Medica xxx (2014) 1e5
References [1] Shirato H, Shimizu S, Kitamura K, Onimaru R. Organ motion in image-guided radiotherapy: lessons from real-time tumor-tracking radiotherapy. Int J Clin Oncol 2007;12:8. [2] Bert C, Durante M. Motion in radiotherapy: particle therapy. Phys Med Biol 2011;56:R113. [3] Schaerer J, Fassi A, Riboldi M, Ceveri P, Baroni G, Sarrut D. Multi-dimensional respiratory motion tracking from markerless optical surface imaging based on deformable mesh registration. Phys Med Biol 2012;57:357. [4] Berbeco RI, Jiang SB, Sharp GC, Chen GT, Mostafavi H, Shirato H. Integrated radiotherapy imaging system (IRIS): design considerations of tumour tracking with linac gantry-mounted diagnostic x-ray systems with flat-panel detectors. Phys Med Biol 2004;49:243. [5] Kupelian P, Willoughby T, Mahadevan A, Djemil T, Weinstein G, Jani S, et al. Multi-institutional clinical experience with the Calypso system in localization and continuous, real-time monitoring of the prostate gland during external radiotherapy. Int J Radiat Oncol Biol Phys 2007;67(4):1088e98. [6] Blanck O, Jauer P, Ernst F, Bruder R, Schweikard A. Pilot-Phantomstudie zur ultraschallgeführten, robotergestützten Radiochirurgie. In: 44. Jahrestagung der Deutschen Gesellschaft für Medizinische Physik (DGMP); 2013. p. 122. [7] Schlosser J, Salisbury K, Xing L, Hritov D. Hybrid X-ray/ultrasound imaging approach for patient positioning and real-time tracking in IGRT. Med Phys 2011;38:3389.
5
[8] Jäckel O, Reiss P. The influence of metal artifacts on the range of ion beams. Phys Med Biol 2007;52:635e44. [9] Saito N, Bert C, Chaudhri N, Gemmel A, Schardt D, Durante M, et al. Speed and accuracy of a beam tracking system for treatment of moving targets with scanned ion beams. Phys Med Biol 2009;54:4849. [10] Zhang X, Günther M, Bongers A. Real-time organ motion tracking in ultrasound imaging using active contours and conditional density propagation. Medical Imaging and Augmented Reality Lecture Notes Comput Sci 2010;6326:286. [11] Kurz C, Bauer J, Bert C, Bongers A, Jenne J, Richter D, et al. First steps towards 4D offline PET-based treatment verification at the Heidelberg Ion Beam Therapy Center. Radiother Oncol 2012;102:55. [12] Tretbar S, Hewener H, Speicher D, Bartscherer T, Bongers A, Jenne J, et al. MR-compatible ultrasound research platform for motion tracking to reduce motion induced artifacts in MR imaging. In: 2013 IEEE International Ultrasonics Symposium (IUS); 2013. p. 553. [13] Fattori G, Saito N, Pella A, Kaderka R, Seregni M, Constantinescu A, et al. TU-ABRA-08: integration of optical tracking for organ motion compensation in scanned ion-beam therapy. Med Phys 2012;39:3889. [14] Chaudhri N, Saito N, Bert C, Franczak B, Steidl B, Durante M, et al. Ion-optical studies for a range adaptation method in ion beam therapy using a static wedge degrader combined with magnetic beam deflection. Phys Med Biol 2010;55:3499e513.
Please cite this article in press as: Schwaab J, et al., Ultrasound tracking for intra-fractional motion compensation in radiation therapy, Physica Medica (2014), http://dx.doi.org/10.1016/j.ejmp.2014.03.003