@article{gaddy_unkelbach_papp_2019, title={Robust spatiotemporal fractionation schemes in the presence of patient setup uncertainty}, volume={46}, ISSN={["2473-4209"]}, url={http://www.scopus.com/inward/record.url?eid=2-s2.0-85067341334&partnerID=MN8TOARS}, DOI={10.1002/mp.13593}, abstractNote={PurposeSpatiotemporal fractionation schemes for photon radiotherapy have recently arisen as a promising technique for healthy tissue sparing. Because spatiotemporally fractionated treatments have a characteristic pattern of delivering high doses to different parts of the tumor in each fraction, uncertainty in patient positioning is an even more pressing concern than in conventional uniform fractionation. Until now, such concerns in patient setup uncertainty have not been addressed in the context of spatiotemporal fractionation.MethodsA stochastic optimization model is used to incorporate patient setup uncertainty to optimize spatiotemporally fractionated plans using expected penalties for deviations from prescription values. First, a robust uniform reference plan is optimized with a stochastic optimization model. Then, a spatiotemporal plan is optimized with a constrained stochastic optimization model that minimizes a primary clinical objective and constrains the spatiotemporal plan to be at least as good as the uniform reference plan with respect to all other objectives. A discrete probability distribution is defined to characterize the random setup error occurring in each fraction. For the optimization of uniform plans, the expected penalties are computed exactly by exploiting the symmetry of the fractions, and for the spatiotemporal plans, quasi‐Monte Carlo sampling is used to approximate the expectation.ResultsUsing five clinical liver cases, it is demonstrated that spatiotemporally fractionated treatment plans maintain the same robust tumor coverage as a stochastic uniform reference plan and exhibit a reduction in the expected mean BED of the uninvolved liver. This is observed for a spectrum of probability distributions of random setup errors with shifts in the patient position of up to 5 mm from the nominal position. For probability distributions with small variance in the patient position, the spatiotemporal plans exhibit an 8–30% reduction in expected mean BED in the healthy liver tissue for shifts up to 2.5 mm and reductions of 5–25% for shifts up to 5 mm.ConclusionsIn the presence of patient setup uncertainty, spatiotemporally fractionated treatment plans exhibit the same robust tumor coverage as their uniformly fractionated counterparts and still retain the benefit in sparing healthy tissues.}, number={7}, journal={MEDICAL PHYSICS}, author={Gaddy, Melissa R. and Unkelbach, Jan and Papp, David}, year={2019}, month={Jul}, pages={2988–3000} } @article{gaddy_yildiz_unkelbach_papp_2018, title={Optimization of spatiotemporally fractionated radiotherapy treatments with bounds on the achievable benefit}, volume={63}, ISSN={["1361-6560"]}, url={http://dx.doi.org/10.1088/1361-6560/aa9975}, DOI={10.1088/1361-6560/aa9975}, abstractNote={Abstract Spatiotemporal fractionation schemes, that is, treatments delivering different dose distributions in different fractions, can potentially lower treatment side effects without compromising tumor control. This can be achieved by hypofractionating parts of the tumor while delivering approximately uniformly fractionated doses to the surrounding tissue. Plan optimization for such treatments is based on biologically effective dose (BED); however, this leads to computationally challenging nonconvex optimization problems. Optimization methods that are in current use yield only locally optimal solutions, and it has hitherto been unclear whether these plans are close to the global optimum. We present an optimization framework to compute rigorous bounds on the maximum achievable normal tissue BED reduction for spatiotemporal plans. The approach is demonstrated on liver tumors, where the primary goal is to reduce mean liver BED without compromising any other treatment objective. The BED-based treatment plan optimization problems are formulated as quadratically constrained quadratic programming (QCQP) problems. First, a conventional, uniformly fractionated reference plan is computed using convex optimization. Then, a second, nonconvex, QCQP model is solved to local optimality to compute a spatiotemporally fractionated plan that minimizes mean liver BED, subject to the constraints that the plan is no worse than the reference plan with respect to all other planning goals. Finally, we derive a convex relaxation of the second model in the form of a semidefinite programming problem, which provides a rigorous lower bound on the lowest achievable mean liver BED. The method is presented on five cases with distinct geometries. The computed spatiotemporal plans achieve 12–35% mean liver BED reduction over the optimal uniformly fractionated plans. This reduction corresponds to 79–97% of the gap between the mean liver BED of the uniform reference plans and our lower bounds on the lowest achievable mean liver BED. The results indicate that spatiotemporal treatments can achieve substantial reductions in normal tissue dose and BED, and that local optimization techniques provide high-quality plans that are close to realizing the maximum potential normal tissue dose reduction.}, number={1}, journal={PHYSICS IN MEDICINE AND BIOLOGY}, author={Gaddy, Melissa R. and Yildiz, Sercan and Unkelbach, Jan and Papp, David}, year={2018}, month={Jan} } @article{gaddy_papp_2016, title={Technical Note: Improving the VMERGE treatment planning algorithm for rotational radiotherapy}, volume={43}, ISSN={["2473-4209"]}, url={http://www.scopus.com/inward/record.url?eid=2-s2.0-84975226392&partnerID=MN8TOARS}, DOI={10.1118/1.4953193}, abstractNote={Purpose:The authors revisit the VMERGE treatment planning algorithm by Craft et al. [“Multicriteria VMAT optimization,” Med. Phys. 39, 686–696 (2012)] for arc therapy planning and propose two changes to the method that are aimed at improving the achieved trade‐off between treatment time and plan quality at little additional planning time cost, while retaining other desirable properties of the original algorithm.Methods:The original VMERGE algorithm first computes an “ideal,” high quality but also highly time consuming treatment plan that irradiates the patient from all possible angles in a fine angular grid with a highly modulated beam and then makes this plan deliverable within practical treatment time by an iterative fluence map merging and sequencing algorithm. We propose two changes to this method. First, we regularize the ideal plan obtained in the first step by adding an explicit constraint on treatment time. Second, we propose a different merging criterion that comprises of identifying and merging adjacent maps whose merging results in the least degradation of radiation dose.Results:The effect of both suggested modifications is evaluated individually and jointly on clinical prostate and paraspinal cases. Details of the two cases are reported.Conclusions:In the authors’ computational study they found that both proposed modifications, especially the regularization, yield noticeably improved treatment plans for the same treatment times than what can be obtained using the original VMERGE method. The resulting plans match the quality of 20‐beam step‐and‐shoot IMRT plans with a delivery time of approximately 2 min.}, number={7}, journal={MEDICAL PHYSICS}, author={Gaddy, Melissa R. and Papp, David}, year={2016}, month={Jul}, pages={4093–4097} }