Room: AAPM ePoster Library
To develop a hybrid direct aperture optimization able to generate treatment plans for VMAT, dynamic trajectory radiotherapy (DTRT) and dynamic mixed beam radiotherapy (DYMBER).
Both DTRT and DYMBER are deliverable on a conventional C-arm treatment unit and both utilize additional degrees of freedom compared to VMAT. DTRT applies dynamic trajectories for photon beams with additional dynamic table and collimator rotation compared to arcs. DYMBER combines photon dynamic trajectories with intensity and energy modulated electron beams collimated with the photon MLC. Both DTRT and DYMBER are non-coplanar treatment techniques. To generate plans for these treatment techniques, we developed a hybrid column generation and simulated annealing algorithm. The column generation algorithm iteratively adds the next promising aperture, either a photon aperture to a predetermined dynamic trajectory or an electron aperture. After each iteration, a combination of gradient descent based MU-weight optimization and simulated annealing based aperture-shape fine-tuning is performed. Deliverable VMAT, DTRT and DYMBER plans are generated with the developed in-house optimizer for two head and neck cases to investigate whether the optimizer can dosimetrically exploit the additional freedom of DTRT and DYMBER compared to VMAT.
The following results are averaged over both cases. Compared to VMAT, D2% of serial OARs is reduced relative to the prescribed dose by 1.1% and 6% for DTRT and DYMBER, respectively and the mean dose of parallel OARs is reduced relative to the prescribed dose by 2.2% and 3.6% for DTRT and DYMBER. Regarding dose homogeneity in the target, D2% / D98% is on average 1.14, 1.13, 1.13 for VMAT, DTRT and DYMBER, respectively.
A hybrid DAO applicable to VMAT, DTRT and DYMBER is successfully developed. It is shown that the optimizer can take dosimetric advantages out of the additional degrees of freedom provided by DTRT and DYMBER compared to VMAT.
Funding Support, Disclosures, and Conflict of Interest: This work was supported by Varian Medical Systems and grant 200021_185366 of the Swiss National Science Foundation.