ESTRO 38 Abstract book

S153 ESTRO 38

stomach. However, the pancreas undergoes erratic and unstable respiratory-induced motion, which makes it challenging to deliver dose accurately. Recent studies have demonstrated that traditional motion-mitigation techniques are often insufficient in guiding SBRT to the pancreas. The purpose of this study was to evaluate real- time imaging and tracking as a tool for pancreatic SBRT and quantify the ability of real-time imaging to correct for unpredictable tumor motion. Material and Methods To understand the effects of tumor motion on treatment, we applies a computational technique that uses the location of implanted fiducial markers in CBCT projection data to reconstruct the motion of these tumors. These data were used to determine the accuracy of 4DCT, and to analyze the impact of respiratory gating, abdominal compression, and real-time imaging on treatment accuracy. Following this, 68 patients were treated with pancreatic SBRT under real-time kV image guidance. Corrections were made to the position if the markers were observed >3 mm from the expected reference position. To understand impact of this imaging on treatment accuracy and clinical workflow, we retrospectively analyzed all treatment interruptions and corrections made based on this imaging. Throughout, we analyzed the effects of tumor motion by developing an artificial neural network-based dosimetric framework to model the dose distribution in the area immediately adjacent to the tumor.

Conclusion The pancreas undergoes unpredictable motion, which decreases the accuracy of 4DCT. Traditional methods of motion mitigation (compression and gating) were unable to decrease the average range of motion below 5 mm. We successfully treated 68 patients using a real-time imaging protocol that corrected positional variations >3 mm, resulting in significant dosimetric benefit to target coverage. In this way, real-time kV imaging allows for the correction of erratic and unpredictable pancreatic motion, and may allow for the safe delivery of future dose- escalated therapies. OC-0302 Dose-guided motion management during liver SBRT delivery using real-time reconstructed tumor DVHs C.G. Muurholm 1 , T. Ravkilde 2 , S. Skouboe 1 , E. Worm 2 , R. Hansen 2 , M. Høyer 3 , P.J. Keall 4 , P.R. Poulsen 1 1 Aarhus University Hospital, Department of Oncology, Aarhus, Denmark; 2 Aarhus University Hospital, Department of Medical Physics, Aarhus, Denmark; 3 Aarhus University Hospital, Danish Center for Particle Therapy, Aarhus, Denmark ; 4 Sydney Medical School - The University Of Sydney, Radiation Physics Laboratory, Sydney, Australia Purpose or Objective Abdominal SBRT is susceptible to tumor motion due to high fraction doses and steep dose gradients. As a QA tool, we have developed a computer program, DoseTracker, that uses streamed tumor positions and linac parameters to reconstruct the dose to a moving tumor in real time during treatment delivery. While the tumor DVH provides intuitive and clinically relevant information on the fully delivered fraction dose it is often not meaningful for a partially delivered fraction. Here, we propose a time- resolved DVH that is useful for partial treatments and demonstrate its use for dose evaluation and dose-guided treatment adaptation in simulated liver SBRT treatments with real-time dose reconstruction. Material and Methods The proposed time-resolved DVH(t) describes the current deviations between the actual delivery with motion and the planned static delivery. Before treatment start DoseTracker calculates the planned static DVH (plan-DVH) of the full treatment. During treatment DoseTracker continuously calculates the cumulative actual dose with motion and planned static dose in all calculation points. The DVH(t) is then obtained by multiplying each calculation point used in the plan-DVH with the ratio of actual and planned dose. Hereby, the full treatment DVH is adjusted based on the current dosimetric state, which results in a real-time calculated DVH that meaningfully describes the current cumulative dose delivery and

Results On average, 4DCT underestimated the 3D range of pancreatic tumor motion by 5.1 mm. These differences were driven by erratic components of respiratory and digestive motion. In retrospective analysis, respiratory gating outperformed abdominal compression, with an average superior-inferior range of motion of 5.5 mm vs 8.5 mm. In the dosimetric model, mean dose errors were less than 5% at all distances from the PTV, and mean absolute dose errors were 5-10%. Real-time imaging resulted in 0.81 pauses per fraction of treatment, with 40% of these resulting in re-localization of the target. The median 3D shift for patient re-alignment was 5.2 mm. 45% of shifts resulted in dosimetric differences to the tumor; of these, the median point dose difference was 23% ± 22% of prescription dose.