Fast on-line plan adjustment for adaptive radiotherapy evaluated for prostate and cervical cancer

Purpose/Objective(s): A new method for fast IMRT plan adjustment based on on-line acquired CT-scans has been developed. Several strategies for using this methodology in daily plan adjustments for prostate and cervical cancer patients have been evaluated. Materials/Methods: Base of the investigations is our 2-phase epsilon-constraint (2pec) multi-criteria optimizer for generation of Pareto-efficient IMRT plans using a prioritized objective list and hard constraints (PMB 54, 7199-7209, PMB 52, 6339-6353). First, for repeat CT-scans of prostate and cervical cancer patients we compared full 2pec optimizations with optimizations using the Lagrange multipliers as obtained for the planning CT, both with respect to plan quality and reduction in calculation time. Next, for groups of these patients, we simulated two adaptive strategies: 1) adaptive therapy using daily plan adjustment while accounting for dose delivered in previous fractions, as assessed with deformable image registration, and 2) adaptive therapy generating for each fraction the best plan of the day, not taking into account previous fractions. In both cases, total delivered doses in the target and OAR were evaluated using deformable image registration. Results: On-line plan generation using Lagrange multipliers as obtained in the full plan optimization for the planning CT-scan is 5- 10 times faster than a full optimization for the repeat scan, depending on the case. For all cases, differences in dose distributions between the fast and full optimization were negligible. For both adaptive strategies, large gains in DVHs for OAR (and sometimes CTV) were observed, compared to the current approach using a single planning CT-scan for all fractions. All OAR DVHs improved in the full dose range. Differences between the strategies in accumulated total dose distributions were virtually absent. Conclusions: Plan adaptation for daily CT-scans using Lagrange multipliers obtained during planning is 5-10 times faster than doing a full daily optimization, while resulting plans are almost identical. Both for prostate and cervical cancer, daily re-planning results in large improvements in accumulated total dose distributions. For the patients studied here, the benefit of daily plan adjustment while accounting for dose delivered in previous fractions is negligible compared to simply generating for each fraction the best plan of the day. As on-line image registration can be omitted, the latter approach is simpler and faster.