Author: Bryan Bednarz, John M. Floberg, Joseph B. Schulz, Jordan M. Slagowski 👨🔬
Affiliation: Department of Medical Physics, School of Medicine and Public Health, University of Wisconsin - Madison, Department of Radiation Oncology, Stanford University School of Medicine, Department of Human Oncology, University of Wisconsin-Madison 🌍
Purpose: Catheter placement for high-dose-rate brachytherapy (HDR BT) is clinician dependent and potentially suboptimal for delivering simultaneous integrated boosts to intraprostatic gross tumor volumes (GTVs) identified on PSMA-PET and/or multiparametric MRI. This study introduces a novel treatment planning system (TPS) that incorporates group sparsity-enforced catheter position optimization, dwell time optimization, and a dwell time homogenizer to maximize dose to the GTV while maintaining overall plan quality.
Methods: An in-house Python-based TPS was developed to integrate clinical DICOM data and simultaneously optimize catheter positions and dose distributions. The system uses the Fast Iterative Shrinkage-Threshold Algorithm to minimize a cost function consisting of a data fidelity term and L2,1 group norm to enforce catheter sparsity. Catheter positions and dwell times are optimized to escalate dose to GTV lesions while enforcing adjacent dwell time homogeneity for robust delivery. The TPS was evaluated retrospectively on nine patient datasets. For each case, the same number of catheters was used in clinical and novel plans, with normalization to ensure a PTV V100% of 95%. Plan quality metrics included GTV D90%, urethra V118%, bladder V80%, and rectum V80%. Statistical significance was assessed using the Wilcoxon signed-rank test.
Results: The proposed optimization technique effectively reduced an average of 76 initial candidate catheter locations to clinical levels (mean of 16 per plan) while improving GTV targeting. GTV D90% increased by 15% (116.2%±6.5% vs 131.2%±11.0%, p=0.004). Dose to critical structures remained within acceptable limits: urethra V118% <0.1cc (0.014±0.019 vs. 0.057±0.032, p=0.014), bladder V80%<1cc (0.25±0.15 vs. 0.83±0.17, p=0.004), rectum<1cc (0.68±0.36 vs. 0.83±0.22, p=0.25).
Conclusion: We demonstrated the feasibility of optimizing HDR BT catheter placement using sparsity-enforced techniques to achieve improved GTV targeting without violating OAR constraints. Future work will focus on real-time implementation for intra-procedural needle guidance.