S148
Brachytherapy - Physics
ESTRO 2026
such optimization is possible in BRIGHT. However, distinctly different trade-offs are obtained compared to dosimetric optimization.
treatment planning is generally guided by population- based dosimetric aims and clinical expertise. Dose- response models estimating Tumor Control Probability (TCP) and Normal Tissue Complication Probability (NTCP) can incorporate additional patient characteristics, allowing for more personalized decision making. Rather than only using these models as decision support, we investigate the potential of the direct integration of such models in cervical cancer brachytherapy treatment plan optimization Material/Methods: We introduce multi-objective optimization using dose- response objectives in BRIGHT [1]. BRIGHT is a multi- center in-house developed AI-based treatment planning method that has been in clinical use for prostate HDR brachytherapy since 2020. BRIGHT performs multi-objective optimization, resulting in a set of treatment plans with different trade-offs between the objectives, generally pertaining to tumor coverage and organ sparing. Ordinarily, dosimetric indices are optimized with respect to EMBRACE-II aims [2]. For dose-response optimization with BRIGHT, we replace the coverage objective by the TCP, and the sparing objective by the average NTCP. TCP and NTCP models used are similar to those in EviGUIDE [3], based on data from the EMBRACE-I study [4]. Endpoints and included predictor variables are shown in labels of Figure 1. Optimization using dosimetric and dose-response objectives is done for 16 patients for which all required data was available, each treated with EBRT+4x combined intracavitary/interstitial brachytherapy. Reported metrics are calculated using 100,000 independently sampled dose calculation points per region of interest. Results: For each patient and each type of optimization objectives, a plan was selected with TCP closest to 95%. Figure 1 shows distributions of patient-wise differences for dosimetric and (N)TCP values when changing from dosimetric to dose-response optimization. This shows a significant reduction of most NTCP values and associated dosimetric indices, though also a significant reduction of CTV_IR D98% for similar TCP values. A key factor is that not all components in the dosimetric coverage objective are in the TCP objective. Figure 2 provides insight into the trade-offs between the objectives for both optimization types. Lower curves indicate lower dosimetric/NTCP values at similar TCP values. For this patient, dosimetric optimization favors lower bladder dose to achieve EMBRACE-II aims, while dose-response optimization favors reduction of the NTCP values for the rectum and recto-vaginal point. Conclusion: Direct multi-objective optimization of dose-response models is an important step towards highly personalized treatment planning. We have shown that
Keywords: dose-response, cervix HDR brachytherapy, BRIGHT References: [1] Dickhoff, L.R.M., et al. "Keeping your best options
Made with FlippingBook - Share PDF online