Interactive GUI for HDR Brachytherapy Planning with Multi-Criteria Optimization
Abstract
Purpose
This work focuses on implementing a GPU-accelerated multi-criteria optimization (MCO) framework for HDR brachytherapy, including advanced applicators such as Direction-Modulated Brachytherapy (DMBT), and developing a graphical user interface (GUI) to enable clinically practical use.
Methods
A GPU-accelerated multi-criteria optimization (MCO) framework for HDR brachytherapy was developed and implemented using PyTorch to generate multiple treatment plans for systematic exploration of clinical trade-offs between target coverage and organ-at-risk sparing. Parallel computation of both optimization and dose calculation was performed on the GPU to enable efficient evaluation of multiple candidate plans. To support advanced and heterogeneous applicators such as Direction-Modulated Brachytherapy (DMBT), a separate optimization workflow was developed to account for applicator-specific dose characteristics and planning constraints. A graphical user interface (GUI) was created using Streamlit to facilitate clinical deployment of the optimization framework. GUI integrates interactive visualization tools for dose distributions, DVHs, and plan selection, enabling efficient and reproducible plan evaluation within a clinically practical workflow.
Results
The multi-criteria optimization framework and graphical user interface were successfully implemented, enabling GPU-accelerated generation of multiple HDR brachytherapy plans through an interactive workflow. The system supports parallel optimization and dose calculation and allows exploration of clinical trade-offs using integrated DVH and dose visualization tools. Initial testing demonstrates computational feasibility and supports a streamlined and reproducible planning workflow.
Conclusion
This work demonstrates the feasibility of a GPU-accelerated, multi-criteria optimization framework for HDR brachytherapy integrated within a graphical user interface to support clinical use. By enabling parallel optimization, interactive trade-off exploration, and support for advanced applicators such as DMBT, the proposed system provides a foundation for efficient brachytherapy treatment planning.