Marc Chamberland

Performance evaluation of real-time motion tracking using positron emission fiducial markers.

PURPOSE Tumor motion due to patient breathing is a factor that limits the accuracy of dose distribution in radiotherapy. One of the methods to improve the accuracy is by applying respiratory gating or tumor tracking. Both techniques require a precise determination of the spatial location of the tumor. We present an experimental evaluation of the performance of PeTrack, a technique that can track internal fiducial markers in real-time for tumor tracking. METHODS PeTrack uses position sensitive detectors to record annihilation coincidence gamma rays from fiducial positron emission markers implanted in or around the tumor. It uses an expectation-maximization clustering algorithm to track the position of the markers. A normalized least mean square adaptive filter was used to predict the position of the markers 100 and 200 ms in the future. We evaluated the performance of the tracking and of the prediction by using a dynamic anthropomorphic thorax phantom to generate three-dimensional (3D) motion of three fiducial markers. The algorithm was run with four different data sets. In the first run, the motion of the markers was based on a sinusoidal model of respiratory motion. Three additional runs were done with motion based on patient breathing data. RESULTS In the case of the sinusoidal model, the average 3D root mean square error for all markers was 0.44 mm. For the three runs based on patient breathing data, the precision of the 3D localization was 0.49 mm. At a latency of 100 ms, the average 3D prediction error was 1.3 +/- 0.6 mm for the sinusoidal model and for the three patient breathing runs. At a latency of 200 ms, the average 3D prediction errors were 1.7 +/- 0.8 mm for the sinusoidal model and 1.4 +/- 0.7 mm for the breathing runs. CONCLUSIONS We conclude that PeTrack can track multiple fiducial markers in real-time with an accuracy and precision smaller than 2 mm. PeTrack can have a direct application in tumor tracking for radiation therapy.

egs_brachy: a versatile and fast Monte Carlo code for brachytherapy.

egs_brachy is a versatile and fast Monte Carlo (MC) code for brachytherapy applications. It is based on the EGSnrc code system, enabling simulation of photons and electrons. Complex geometries are modelled using the EGSnrc C++ class library and egs_brachy includes a library of geometry models for many brachytherapy sources, in addition to eye plaques and applicators. Several simulation efficiency enhancing features are implemented in the code. egs_brachy is benchmarked by comparing TG-43 source parameters of three source models to previously published values. 3D dose distributions calculated with egs_brachy are also compared to ones obtained with the BrachyDose code. Well-defined simulations are used to characterize the effectiveness of many efficiency improving techniques, both as an indication of the usefulness of each technique and to find optimal strategies. Efficiencies and calculation times are characterized through single source simulations and simulations of idealized and typical treatments using various efficiency improving techniques. In general, egs_brachy shows agreement within uncertainties with previously published TG-43 source parameter values. 3D dose distributions from egs_brachy and BrachyDose agree at the sub-percent level. Efficiencies vary with radionuclide and source type, number of sources, phantom media, and voxel size. The combined effects of efficiency-improving techniques in egs_brachy lead to short calculation times: simulations approximating prostate and breast permanent implant (both with (2 mm)3 voxels) and eye plaque (with (1 mm)3 voxels) treatments take between 13 and 39 s, on a single 2.5 GHz Intel Xeon E5-2680 v3 processor core, to achieve 2% average statistical uncertainty on doses within the PTV. egs_brachy will be released as free and open source software to the research community.

Algorithm and simulation for real-time positron emission based tumor tracking using a linear fiducial marker

The effectiveness of radiotherapy in cancer treatment remains significantly limited by the accuracy of tumor dose delivery. The ideal solution lies in real-time localization of patient tumors during therapy; one such method is by tracking implanted low-activity positron emitters using two pairs of orthogonally placed gamma-ray detectors. Prior studies have examined multiple point sources, which have potential patient complications during implantation. A linear source geometry is proposed as a less invasive alternative, with potential higher-precision tracking. A source localization algorithm has been devised using cost-function minimization of the source position estimate relative to annihilation gamma coincidence lines. The algorithm was tested via Monte Carlo simulation methods using a Geant4 application for emission tomography (GATE) package for a source of length of 2.00 cm and width of 0.1 mm. The midpoint of the linear marker was located within submillimeter accuracy at 200 coincidence events and the orientation of the source determined with less than 5 degrees (0.087 rad) angular deviation at 300 events. At an optimal event count of 700, tracking had mean midpoint error of 0.48 +/- 0.26 mm and mean angular deviation of 0.041 +/- 0.023 rad (1.4 degrees +/- 0.8 degree). The source and tracking algorithm may prove effective for future clinical implementation in radiotherapy treatment.

A generic TG-186 shielded applicator for commissioning model-based dose calculation algorithms for high-dose-rate 192Ir brachytherapy

Purpose: A joint working group was created by the American Association of Physicists in Medicine (AAPM), the European Society for Radiotherapy and Oncology (ESTRO), and the Australasian Brachytherapy Group (ABG) with the charge, among others, to develop a set of well‐defined test case plans and perform calculations and comparisons with model‐based dose calculation algorithms (MBDCAs). Its main goal is to facilitate a smooth transition from the AAPM Task Group No. 43 (TG‐43) dose calculation formalism, widely being used in clinical practice for brachytherapy, to the one proposed by Task Group No. 186 (TG‐186) for MBDCAs. To do so, in this work a hypothetical, generic high‐dose rate (HDR) 192Ir shielded applicator has been designed and benchmarked. Methods: A generic HDR 192Ir shielded applicator was designed based on three commercially available gynecological applicators as well as a virtual cubic water phantom that can be imported into any DICOM‐RT compatible treatment planning system (TPS). The absorbed dose distribution around the applicator with the TG‐186 192Ir source located at one dwell position at its center was computed using two commercial TPSs incorporating MBDCAs (Oncentra® Brachy with Advanced Collapsed‐cone Engine, ACE™, and BrachyVision ACUROS™) and state‐of‐the‐art Monte Carlo (MC) codes, including ALGEBRA, BrachyDose, egs_brachy, Geant4, MCNP6, and Penelope2008. TPS‐based volumetric dose distributions for the previously reported “source centered in water” and “source displaced” test cases, and the new “source centered in applicator” test case, were analyzed here using the MCNP6 dose distribution as a reference. Volumetric dose comparisons of TPS results against results for the other MC codes were also performed. Distributions of local and global dose difference ratios are reported. Results: The local dose differences among MC codes are comparable to the statistical uncertainties of the reference datasets for the “source centered in water” and “source displaced” test cases and for the clinically relevant part of the unshielded volume in the “source centered in applicator” case. Larger local differences appear in the shielded volume or at large distances. Considering clinically relevant regions, global dose differences are smaller than the local ones. The most disadvantageous case for the MBDCAs is the one including the shielded applicator. In this case, ACUROS agrees with MC within [−4.2%, +4.2%] for the majority of voxels (95%) while presenting dose differences within [−0.12%, +0.12%] of the dose at a clinically relevant reference point. For ACE, 95% of the total volume presents differences with respect to MC in the range [−1.7%, +0.4%] of the dose at the reference point. Conclusions: The combination of the generic source and generic shielded applicator, together with the previously developed test cases and reference datasets (available in the Brachytherapy Source Registry), lay a solid foundation in supporting uniform commissioning procedures and direct comparisons among treatment planning systems for HDR 192Ir brachytherapy.

TU-AB-BRC-08: Egs_brachy, a Fast and Versatile Monte Carlo Code for Brachytherapy Applications

PURPOSE To introduce egs_brachy, a new, fast, and versatile Monte Carlo code for brachytherapy applications. METHODS egs_brachy is an EGSnrc user-code based on the EGSnrc C++ class library (egs++). Complex phantom, applicator, and source model geometries are built using the egs++ geometry module. egs_brachy uses a tracklength estimator to score collision kerma in voxels. Interaction, spectrum, energy fluence, and phase space scoring are also implemented. Phase space sources and particle recycling may be used to improve simulation efficiency. HDR treatments (e.g. stepping source through dwell positions) can be simulated. Standard brachytherapy seeds, as well as electron and miniature x-ray tube sources are fully modelled. Variance reduction techniques for electron source simulations are implemented (Bremsstrahlung cross section enhancement, uniform Bremsstrahlung splitting, and Russian Roulette). TG-43 parameters of seeds are computed and compared to published values. Example simulations of various treatments are carried out on a single 2.5 GHz Intel Xeon E5-2680 v3 processor core. RESULTS TG-43 parameters calculated with egs_brachy show excellent agreement with published values. Using a phase space source, 2% average statistical uncertainty in the PTV ((2mm)3 voxels) can be achieved in 10 s for 100 125 I or 103 Pd seeds in a 36.2 cm3 prostate PTV, 31 s for 64 103 Pd seeds in a 64 cm3 breast PTV, and 56 s for a miniature x-ray tube in a 27 cm3 breast PTV. Comparable uncertainty is reached in 12 s in a (1 mm)3 water voxel 5 mm away from a COMS 16mm eye plaque with 13 103 Pd seeds. CONCLUSION The accuracy of egs_brachy has been demonstrated through benchmarking calculations. Calculation times are sufficiently fast to allow full MC simulations for routine treatment planning for diverse brachytherapy treatments (LDR, HDR, miniature x-ray tube). egs_brachy will be available as free and open-source software to the medical physics research community. This work is partially funded by the Canada Research Chairs program, the Natural Sciences and Engineering Research Council of Canada, and the Ontario Ministry of Research and Innovation (Ontario Early Researcher Award).

Evaluation of the clinical efficacy of the PeTrack motion tracking system for respiratory gating in cardiac PET imaging

Respiratory gating is a common technique used to compensate for patient breathing motion and decrease the prevalence of image artifacts that can impact diagnoses. In this study a new data-driven respiratory gating method (PeTrack) was compared with a conventional optical tracking system. The performance of respiratory gating of the two systems was evaluated by comparing the number of respiratory triggers, patient breathing intervals and gross heart motion as measured in the respiratory-gated image reconstructions of rubidium-82 cardiac PET scans in test and control groups consisting of 15 and 8 scans, respectively. We found evidence suggesting that PeTrack is a robust patient motion tracking system that can be used to retrospectively assess patient motion in the event of failure of the conventional optical tracking system.

Reply to Comment on ‘egs_brachy: a versatile and fast Monte Carlo code for brachytherapy’

We respond to the comments by Dr Yegin by identifying the source of an error in a fit in our original paper but arguing that the lack of a fit does not affect the conclusion based on the raw data that [Formula: see text] is an accurate code and we provide further benchmarking data to demonstrate this point.

Sci-Thur PM – Brachytherapy 01: Fast brachytherapy dose calculations: Characterization of egs_brachy features to enhance simulation efficiency

Purpose: egs_brachy is a fast, new EGSnrc user-code for brachytherapy applications. This study characterizes egs_brachy features that enhance simulation efficiency. Methods: Calculations are performed to characterize efficiency gains from various features. Simulations include radionuclide and miniature x-ray tube sources in water phantoms and idealized prostate, breast, and eye plaque treatments. Features characterized include voxel indexing of sources to reduce boundary checks during radiation transport, scoring collision kerma via tracklength estimator, recycling photons emitted from sources, and using phase space data to initiate simulations. Bremsstrahlung cross section enhancement (BCSE), uniform bremsstrahlung splitting (UBS), and Russian Roulette (RR) are considered for electronic brachytherapy. Results: Efficiency is enhanced by a factor of up to 300 using tracklength versus interaction scoring of collision kerma and by up to 2.7 and 2.6 using phase space sources and particle recycling respectively compared to simulations in which particles are initiated within sources. On a single 2.5 GHz Intel Xeon E5-2680 processor cor, simulations approximating prostate and breast permanent implant ((2 mm)3 voxels) and eye plaque ((1 mm)3) treatments take as little as 9 s (prostate, eye) and up to 31 s (breast) to achieve 2% statistical uncertainty on doses within the PTV. For electronic brachytherapy, BCSE, UBS, and RR enhance efficiency by a factor >2000 compared to a factor of >104 using a phase space source. Conclusion: egs_brachy features provide substantial efficiency gains, resulting in calculation times sufficiently fast for full Monte Carlo simulations for routine brachytherapy treatment planning.

Technical aspects of real time positron emission tracking for gated radiotherapy.

PURPOSE Respiratory motion can lead to treatment errors in the delivery of radiotherapy treatments. Respiratory gating can assist in better conforming the beam delivery to the target volume. We present a study of the technical aspects of a real time positron emission tracking system for potential use in gated radiotherapy. METHODS The tracking system, called PeTrack, uses implanted positron emission markers and position sensitive gamma ray detectors to track breathing motion in real time. PeTrack uses an expectation-maximization algorithm to track the motion of fiducial markers. A normalized least mean squares adaptive filter predicts the location of the markers a short time ahead to account for system response latency. The precision and data collection efficiency of a prototype PeTrack system were measured under conditions simulating gated radiotherapy. The lung insert of a thorax phantom was translated in the inferior-superior direction with regular sinusoidal motion and simulated patient breathing motion (maximum amplitude of motion ±10 mm, period 4 s). The system tracked the motion of a (22)Na fiducial marker (0.34 MBq) embedded in the lung insert every 0.2 s. The position of the was marker was predicted 0.2 s ahead. For sinusoidal motion, the equation used to model the motion was fitted to the data. The precision of the tracking was estimated as the standard deviation of the residuals. Software was also developed to communicate with a Linac and toggle beam delivery. In a separate experiment involving a Linac, 500 monitor units of radiation were delivered to the phantom with a 3 × 3 cm photon beam and with 6 and 10 MV accelerating potential. Radiochromic films were inserted in the phantom to measure spatial dose distribution. In this experiment, the period of motion was set to 60 s to account for beam turn-on latency. The beam was turned off when the marker moved outside of a 5-mm gating window. RESULTS The precision of the tracking in the IS direction was 0.53 mm for a sinusoidally moving target, with an average count rate ∼250 cps. The average prediction error was 1.1 ± 0.6 mm when the marker moved according to irregular patient breathing motion. Across all beam deliveries during the radiochromic film measurements, the average prediction error was 0.8 ± 0.5 mm. The maximum error was 2.5 mm and the 95th percentile error was 1.5 mm. Clear improvement of the dose distribution was observed between gated and nongated deliveries. The full-width at halfmaximum of the dose profiles of gated deliveries differed by 3 mm or less than the static reference dose distribution. Monitoring of the beam on/off times showed synchronization with the location of the marker within the latency of the system. CONCLUSIONS PeTrack can track the motion of internal fiducial positron emission markers with submillimeter precision. The system can be used to gate the delivery of a Linac beam based on the position of a moving fiducial marker. This highlights the potential of the system for use in respiratory-gated radiotherapy.

List-mode motion tracking for positron emission tomography imaging using low-activity fiducial markers

Motion correction in positron emission tomography (PET) imaging may benefit from an accurate knowledge of the patient motion during the image acquisition. We evaluated the feasibility to detect and record the motion of external fiducial markers during PET scans. We developed an iterative algorithm that can track the three-dimensional (3D) motion of low-activity fiducial markers while surrounded by high physiological tracer activity from a patient undergoing PET imaging. Monte Carlo techniques were used to simulate a 92.5-kBq 22Na marker moving sinusoidally in 3D. The simulated events were combined with list-mode data from patients undergoing cardiac PET imaging in order to test the algorithm. In experimental studies, three external 22Na markers were placed on a dynamic torso phantom with an initial activity of approximately 680 MBq of 82Rb in its cardiac insert. We tracked the motion of those markers while simulating breathing motion and patient drift with the phantom. Results from simulations show that a 92.5-kBq marker can be tracked in 3D at a frequency of 2 Hz with an accuracy of 1.2 mm and a precision of 0.8 mm. The phantom study qualitatively confirms that the algorithm can track both breathing and patient motion. The relative accuracy of the tracking was 0.4±1.1 mm and the precision was 0.8 mm. In conclusion, we have developed an algorithm that can track the 3D motion of low-activity positron-emitting markers during PET imaging. This motion information might prove useful in developing new motion correction schemes in clinical PET imaging.