ESTRO 2021 Abstract Book

S1482

ESTRO 2021

dose in real time. This will enable dose-guided treatment adaptation and will provide confidence for the implementation of real-time adapted radiotherapy. Materials and Methods Our workflow comprises three steps. First, a synthetic CT (sCT) is created from the daily 3D MR acquired directly prior to treatment. Electron densities and material parameters are assigned to anatomical structures using indication specific look-up tables. Second, motion induced displacement vectors are determined for the PTV using a template matching algorithm. A cine image (2Dref) is selected and a template outline is created from a mask of the PTV. The corresponding slice of the 3D daily MR image in the plane of 2Dref and their relative displacement is evaluated. Subsequent cine images are compared to the 2Dref template and the sum of the 2D-2Dref and 3D- 2Dref displacements is inverted to yield the isocentre shift vector. Finally, beam delivery parameters are read from the linac, the isocentre shift is applied, and the delivered dose is calculated on the sCT using a research version of the Monte Carlo algorithm we use in the clinic (Hissoiny et al). A small number of particles are simulated repeatedly to determine the PTV-motion-including segment dose. The resulting segment doses are averaged and accumulated to the total delivered dose. A dose calculation excluding the effect of motion is carried out in parallel. Motion-including and excluding dose volume histogram constraints (DVHc) are calculated at a frequency of 1Hz during treatment. A prostate patient (60Gy/20#) was selected for real time dose reconstruction during treatment delivery. A further five prostate patients (36.25Gy/5#) were selected for dose reconstruction in a simulation environment. Calculations used 2 mm cubic voxels on a desktop computer with an Intel Xenon E5 processor and 32GB of RAM. Results We successfully reconstructed motion including dose in real time during treatment of the 60Gy patient. The patient exhibited minimal motion, therefore changes to DVHc were small. Data from the simulated case with the largest PTV motion is presented in figure 1. During treatment, the CTV moved outside the 3mm PTV margin, reducing CTV coverage.

During real time dose reconstruction, the calculation uncertainty (σ) decreases as more particles are simulated. Figure 2 shows its variation for the case presented above.