Abstract
Introducing anatomical changes to a patient during intensity-modulated proton therapy (IMPT) can cause inadequate target coverage and overdosage to normal tissues. Although adaptive radiation therapy (ART) has been proposed to address this issue, the lack of integrated automated systems has limited its adoption. The objective of this study was to develop and validate image-guided adaptive strategies for robustly optimized IMPT (RO-IMPT) using cone-beam computed tomography (CBCT) and an automated planning system. The impact of uncertainty in RO-IMPT was first investigated and found that reducing setup uncertainty has more benefits than reducing range uncertainty for head and neck (HN) patients. To achieve CBCT-based ART strategy, this study validated the use of deformable image registration (DIR) for CT-to-CBCT registration and proposed an automated approach for evaluating DIR quality. CBCT-based proton dose evaluation by employing DIR was then investigated, indicating that the uncertainty caused by DIR errors was far smaller than the dose variation stemming from anatomical changes. To facilitate replanning, proton-specific knowledge-based planning models were trained and validated for automatic prostate and HN IMPT plan generation. The feasibility and potential benefits of daily adaptation using these strategies were evaluated for HN cancer patients, showing improved target coverage, OAR sparing, and lower normal tissue complication probabilities compared to non-adaptive IMPT. In conclusion, this study demonstrated the feasibility and potential benefits of adaptive strategies for patients undergoing RO-IMPT treatment based on pretreatment CBCT and an automated planning model.