- Email: [email protected]
Real-time Tumor Position Monitoring and Dynamic Dose Adaptation: Geometric and Dosimetric Accuracy of an Integrated Tracking System
I. J. Radiation Oncology d Biology d Physics
S74
Volume 75, Number 3, Supplement, 2009
the intrinsic dosimetric inequality even among neighboring voxels, the prescription is chosen uniformly for all voxels within a structure. Therefore, the hidden potential of personalized voxel-specific prescription is not exploited in the conventional inverse planning. Here we introduce the idea of iterative adjustment of voxel-specific prescription doses. This way there is no need to prespecify the customized prescription; it is determined automatically as the result of iterative adjustments. Our method can be summarized as follows. We start by solving the optimization problem for the conventional uniform prescription. Using the resulting optimal fluence map we calculate the dose delivered to each voxel. We then compare it to the prescribed dose. If a voxel is over(under-) dosed then its prescription dose for the next iteration is decreased (increased) and the plan is reoptimized. In other words both under- and over-dosed voxels are penalized with a penalty weight proportional to the amount of the dose departure from a current prescription. The procedure is repeated until the voxel dose cannot be improved anymore. Results: A clinical head-and-neck case was used to test our method. By adjusting voxel prescriptions to compensate for the inequalities between the actual calculated and desired doses, substantial improvements are obtained for the treatment plan as large dose reductions were achieved in almost all of the critical structures present comparing to the conventional uniform prescription plan with the same PTV coverage. For instance, we demonstrate fivefold reduction in the maximum dose to the brainstem. Other organs at risk experience dose reduction ranging from 100–300 percent. Conclusions: We demonstrated that our method of prescription adjustment is capable of producing much better dose distributions in organs at risk, iteratively, without expert’s intervention. The method can readily be implemented with any treatment planning system and demonstrates fast convergence and significantly improved plans. Author Disclosure: P. Lougovski, None; J. LeNoach, None; L. Zhu, None; Y. Ma, None; L. Xing, None.
156
Real-time Tumor Position Monitoring and Dynamic Dose Adaptation: Geometric and Dosimetric Accuracy of an Integrated Tracking System
A. Krauss, A. W. Rau, M. Tacke, S. Nill, U. Oelfke German Cancer Research Center (DKFZ), Heidelberg, Germany Purpose/Objective(s): Dynamic dose adaptation aims to efficiently compensate for dose delivery errors due to intrafraction organ motion. We developed a 4D-tumor tracking platform that combines a real-time tumor localization system with a continuous data feed-back loop from the dose delivery process to adapt the leaf positions of a dynamic MLC to the observed organ motion. The performance of the system is investigated in experiments with 2D motion patterns of phantoms. Characteristic parameters like system latencies, geometric accuracy and dosimetric accuracy are determined. Materials/Methods: Real-time tumor position monitoring, provided by the electromagnetic tracking technology of Calypso Medical Technologies, Inc, with an update rate of approximately 25 Hz (a research version with faster update rate) was utilized to dynamically adapt the aperture of the radiation field delivered by a MLC-160 of a Siemens Artiste. The MLC was driven by an investigational leaf control system. The phantom consisted of a stack of water-equivalent slices (RW3) with three electromagnetic transponders embedded and positioned on a 2D mobile stage. The dosimetric evaluation was based on radiochromic films placed inside the phantom. The position of the phantom was simultaneously monitored by the electromagnetic system and by two potentiometers (for longitudinal/lateral) digitized at 300 Hz. In a first experiment, a sinusoidal motion pattern with amplitude of 2 cm and a repetition rate of 4 seconds was used for the longitudinal direction (LD) whereas a reduced amplitude of 1 cm and a repetition rate of 5.4 seconds were used for the lateral motion. The leaf travel direction was parallel to the longitudinal axis for a set of open square fields. The geometric accuracy was assessed by capturing portal images at 15 Hz showing both the current phantom position and the current MLC aperture. Results: The latency of the integrated system was preliminarily estimated to be 500 ms, predominantly caused by the MLC control unit. A second order prediction algorithm applied to the target motion allowed accurate tracking of the experimental 2D target motion. Film dosimetry yielded penumbras along the leaves direction of 4 mm, 15 mm, and 7 mm in the cases of no motion, motion without and with MLC tracking, respectively. Analysis of the dose patterns with a gamma index criterion of 2%/2 mm revealed an increase of the success rate from 39–85% for 2D-tracking and from 47–93% for 1D-tracking along the leaves direction. Conclusions: The feasibility of integrated real-time tumor tracking using a combination of the Calypso System and an MLC-160 operating in dynamic mode was demonstrated by phantom experiments. Author Disclosure: A. Krauss, None; A.W. Rau, None; M. Tacke, None; S. Nill, None; U. Oelfke, None.
157
PET Registration Methods and Dosimetric Consequences
G. J. Kubicek1, S. Fogh1, J. W. Piper2, Y. Xiao1, M. Machtay1 1 Thomas Jefferson University Hospital, Philadelphia, PA, 2Department of Computer Science, Wake Forest University, Winston-Salem, NC
Purpose/Objective(s): The PET/CT provides valuable information for head-and-neck cancer (HNC) radiation treatment planning (RTP). Ideally, a dedicated RTP PET/CT should be performed for RTP, however, this is often not logistically or financially feasible. Methods to incorporate previously obtained PET/CT data include rigid registration techniques (RRT) and deformable registration techniques (DRT). The DRT uses a constrained intensity-based free-form deformable CT-CT registration algorithm allowing for the diagnostic CT scan to be deformed in order to match the neck position of the planning CT scan. We evaluated the outcome of different PET scan fusion techniques. Materials/Methods: We retrospectively evaluated 10 head-and-neck cancer patients that were planned and treated without PET fusion. The GTV and PTVs were defined as the actual GTV and PTV used for radiotherapy contoured from nonfused PET scans and clinical data. These clinical contours were compared to PET-derived GTVs obtained via RRT or DRT techniques using two PET methods: maximal SUV (3.0) and ‘PET edge’ techniques. Thus, a total of 5 GTVs were obtained for each patient. Fusion accuracy was measured by center of mass (COM) and overlap differences between the PET derived and planning GTVs and also the coverage difference between the PET derived GTVs and the PTV.
S74
Volume 75, Number 3, Supplement, 2009
the intrinsic dosimetric inequality even among neighboring voxels, the prescription is chosen uniformly for all voxels within a structure. Therefore, the hidden potential of personalized voxel-specific prescription is not exploited in the conventional inverse planning. Here we introduce the idea of iterative adjustment of voxel-specific prescription doses. This way there is no need to prespecify the customized prescription; it is determined automatically as the result of iterative adjustments. Our method can be summarized as follows. We start by solving the optimization problem for the conventional uniform prescription. Using the resulting optimal fluence map we calculate the dose delivered to each voxel. We then compare it to the prescribed dose. If a voxel is over(under-) dosed then its prescription dose for the next iteration is decreased (increased) and the plan is reoptimized. In other words both under- and over-dosed voxels are penalized with a penalty weight proportional to the amount of the dose departure from a current prescription. The procedure is repeated until the voxel dose cannot be improved anymore. Results: A clinical head-and-neck case was used to test our method. By adjusting voxel prescriptions to compensate for the inequalities between the actual calculated and desired doses, substantial improvements are obtained for the treatment plan as large dose reductions were achieved in almost all of the critical structures present comparing to the conventional uniform prescription plan with the same PTV coverage. For instance, we demonstrate fivefold reduction in the maximum dose to the brainstem. Other organs at risk experience dose reduction ranging from 100–300 percent. Conclusions: We demonstrated that our method of prescription adjustment is capable of producing much better dose distributions in organs at risk, iteratively, without expert’s intervention. The method can readily be implemented with any treatment planning system and demonstrates fast convergence and significantly improved plans. Author Disclosure: P. Lougovski, None; J. LeNoach, None; L. Zhu, None; Y. Ma, None; L. Xing, None.
156
Real-time Tumor Position Monitoring and Dynamic Dose Adaptation: Geometric and Dosimetric Accuracy of an Integrated Tracking System
A. Krauss, A. W. Rau, M. Tacke, S. Nill, U. Oelfke German Cancer Research Center (DKFZ), Heidelberg, Germany Purpose/Objective(s): Dynamic dose adaptation aims to efficiently compensate for dose delivery errors due to intrafraction organ motion. We developed a 4D-tumor tracking platform that combines a real-time tumor localization system with a continuous data feed-back loop from the dose delivery process to adapt the leaf positions of a dynamic MLC to the observed organ motion. The performance of the system is investigated in experiments with 2D motion patterns of phantoms. Characteristic parameters like system latencies, geometric accuracy and dosimetric accuracy are determined. Materials/Methods: Real-time tumor position monitoring, provided by the electromagnetic tracking technology of Calypso Medical Technologies, Inc, with an update rate of approximately 25 Hz (a research version with faster update rate) was utilized to dynamically adapt the aperture of the radiation field delivered by a MLC-160 of a Siemens Artiste. The MLC was driven by an investigational leaf control system. The phantom consisted of a stack of water-equivalent slices (RW3) with three electromagnetic transponders embedded and positioned on a 2D mobile stage. The dosimetric evaluation was based on radiochromic films placed inside the phantom. The position of the phantom was simultaneously monitored by the electromagnetic system and by two potentiometers (for longitudinal/lateral) digitized at 300 Hz. In a first experiment, a sinusoidal motion pattern with amplitude of 2 cm and a repetition rate of 4 seconds was used for the longitudinal direction (LD) whereas a reduced amplitude of 1 cm and a repetition rate of 5.4 seconds were used for the lateral motion. The leaf travel direction was parallel to the longitudinal axis for a set of open square fields. The geometric accuracy was assessed by capturing portal images at 15 Hz showing both the current phantom position and the current MLC aperture. Results: The latency of the integrated system was preliminarily estimated to be 500 ms, predominantly caused by the MLC control unit. A second order prediction algorithm applied to the target motion allowed accurate tracking of the experimental 2D target motion. Film dosimetry yielded penumbras along the leaves direction of 4 mm, 15 mm, and 7 mm in the cases of no motion, motion without and with MLC tracking, respectively. Analysis of the dose patterns with a gamma index criterion of 2%/2 mm revealed an increase of the success rate from 39–85% for 2D-tracking and from 47–93% for 1D-tracking along the leaves direction. Conclusions: The feasibility of integrated real-time tumor tracking using a combination of the Calypso System and an MLC-160 operating in dynamic mode was demonstrated by phantom experiments. Author Disclosure: A. Krauss, None; A.W. Rau, None; M. Tacke, None; S. Nill, None; U. Oelfke, None.
157
PET Registration Methods and Dosimetric Consequences
G. J. Kubicek1, S. Fogh1, J. W. Piper2, Y. Xiao1, M. Machtay1 1 Thomas Jefferson University Hospital, Philadelphia, PA, 2Department of Computer Science, Wake Forest University, Winston-Salem, NC
Purpose/Objective(s): The PET/CT provides valuable information for head-and-neck cancer (HNC) radiation treatment planning (RTP). Ideally, a dedicated RTP PET/CT should be performed for RTP, however, this is often not logistically or financially feasible. Methods to incorporate previously obtained PET/CT data include rigid registration techniques (RRT) and deformable registration techniques (DRT). The DRT uses a constrained intensity-based free-form deformable CT-CT registration algorithm allowing for the diagnostic CT scan to be deformed in order to match the neck position of the planning CT scan. We evaluated the outcome of different PET scan fusion techniques. Materials/Methods: We retrospectively evaluated 10 head-and-neck cancer patients that were planned and treated without PET fusion. The GTV and PTVs were defined as the actual GTV and PTV used for radiotherapy contoured from nonfused PET scans and clinical data. These clinical contours were compared to PET-derived GTVs obtained via RRT or DRT techniques using two PET methods: maximal SUV (3.0) and ‘PET edge’ techniques. Thus, a total of 5 GTVs were obtained for each patient. Fusion accuracy was measured by center of mass (COM) and overlap differences between the PET derived and planning GTVs and also the coverage difference between the PET derived GTVs and the PTV.