[OA011] DoseTracker: In-house developed software program for real-time reconstruction of motion-induced dose errors during radiotherapy

Purpose A problem of current plan-specific quality assurance (QA) in radiotherapy is that it ignores organ motion, although motion may cause dose discrepancies much graver than the treatment delivery errors assessed by the QA procedure. Furthermore, the QA is not feasible for real-time adaptive treatments like MLC tracking, where the treatment machine behavior is determined on-the-fly as a response to tumor motion. To overcome these limitations, we have developed the software program DoseTracker that performs real-time motion-including dose reconstruction during treatment delivery. Here, we validate DoseTracker in experiments and simulated treatments. Methods DoseTracker performs real-time motion-including dose reconstruction, based on streamed linac parameters and target positions, using a pencil-beam algorithm. An arbitrary set of points can be chosen for the dose calculations and the calculations points can move independently as function of time. DoseTracker has been validated in phantom experiments and simulated patient treatments.The phantom experiments used a biplanar diode array on a programmable motion stage. DoseTracker performed online real-time reconstruction of the diode doses at 15 Hz and calculated the 3%/3 mm gamma failure rate (comparing motion doses with static doses) at 1 Hz during five VMAT liver SBRT treatments with and without MLC tracking. The gamma errors were retrospectively compared with diode measurements.In the treatment simulations, DoseTracker made (offline) real-time tumor dose reconstructions at 2–5 Hz in patient anatomy for four liver SBRT patients previously treated with motion monitoring by electromagnetic transponders. Simulations were performed with and without respiratory gating. The reduction in CTV D95 relative to the planned intent (ΔD95) was retrospectively compared between DoseTracker and motion-including dose reconstructions performed in the treatment planning system (TPS) by isocenter shifts. DoseTracker currently assumes water densities so another set of TPS calculations were performed on water densities. Results Experiments yielded 2.0%-point gamma failure rate root-mean-square difference between DoseTracker and measurements. Simulated treatments yielded 1.2%-points (CT-densities) and 0.6%-point (water-densities) differences in ΔD95. Conclusions A program to reconstruct motion-induced dose errors was developed and tested in phantom studies and patients, yielding high accuracies, and allowing supervision of treatment correctness and action based on dose discrepancies. Planned improvements include CT-densities, rotations, etc.