[論文レビュー] OpenTPS -- Open-source treatment planning system for research in proton therapy

S. Wuyckens, D. Dasnoy+10 more|arXiv (Cornell University)|Mar 1, 2023

Radiation Therapy and Dosimetry被引用数 9

ひとこと要約

OpenTPSは、プロトン治療研究のためのオープンソースのPythonベースの治療計画システムで、柔軟なプラン作成、線量計算、堅牢な最適化をコミュニティ主導の開発で実現します。

ABSTRACT

Introduction. Treatment planning systems (TPS) are an essential component for simulating and optimizing a radiation therapy treatment before administering it to the patient. It ensures that the tumor is well covered and the dose to the healthy tissues is minimized. However, the TPS provided by commercial companies often come with a large panel of tools, each implemented in the form of a black-box making it difficult for researchers to use them for implementing and testing new ideas. To address this issue, we have developed an open-source TPS. Approach. We have developed an open-source software platform, OpenTPS (opentps.org), to generate treatment plans for external beam radiation therapy, and in particular for proton therapy. It is designed to be a flexible and user-friendly platform (coded with the freely usable Python language) that can be used by medical physicists, radiation oncologists, and other members of the radiation therapy community to create customized treatment plans for educational and research purposes. Result. OpenTPS includes a range of tools and features that can be used to analyze patient anatomy, simulate the delivery of the radiation beam, and optimize the treatment plan to achieve the desired dose distribution. It can be used to create treatment plans for a variety of cancer types and was designed to be extended to other treatment modalities. Significance. A new open-source treatment planning system has been built for research in proton therapy. Its flexibility allows an easy integration of new techniques and customization of treatment plans. It is freely available for use and is regularly updated and supported by a community of users and developers who contribute to the ongoing development and improvement of the software.

研究の動機と目的

• Motivate and enable research in proton therapy by providing an openly accessible TPS platform.
• Offer a flexible, extensible software architecture (Core and GUI) for data management, dose computation, planning, and evaluation.
• Facilitate integration of new dose engines, optimization methods, and robustness concepts within a shared framework.
• Support educational use and AI/machine learning research through data augmentation and 4D imaging tools.

提案手法

• Develop a two-package architecture: Core library for data handling, dose computation, and planning; GUI for visualization and interaction.
• Integrate fast Monte Carlo dose calculation via MCsquare to compute beamlet doses and dose distributions.
• Implement IMPT-focused plan design and optimization using voxel-based dose modeling with a beamlet matrix A.
• Provide multiple optimization solvers (gradient-based, quasi-Newton, FISTA, LP, interior point, with Gurobi as an option).
• Incorporate robust optimization using worst-case scenario evaluation across multiple setup and range uncertainties.
• Support robust and 4D dose simulations (4DCT, MidP, Deformation Fields) and intra-/inter-fraction motion analyses.
• Include data augmentation tools to simulate inter- and intra-fraction motions for AI/data generation.

Figure 1: Main window of the OpenTPS GUI.

実験結果

リサーチクエスチョン

• RQ1 How can an open-source TPS support flexible, research-oriented development in proton therapy?
• RQ2 What optimization strategies and solvers are effective for IMPT plan optimization within an open framework?
• RQ3 How can robustness to setup, range, and motion uncertainties be incorporated and evaluated in a research TPS?
• RQ4 How can 4D dose and motion models be integrated and used for plan evaluation and data generation?
• RQ5 Can an open platform accelerate integration of new modalities (e.g., arc therapy) and AI-assisted planning?

主な発見

• OpenTPS provides a modular Core/GUI architecture with data handling, dose calculation (MCsquare), and plan optimization capabilities.
• It supports a range of solvers (gradient-based, FISTA, LBFGS, LP, interior point) and a Gurobi-based linear model for optimization.
• Robust optimization using worst-case scenarios across 21 scenarios (7 setup x 3 range) is implemented and operable.
• 4D dose simulations (4DD and 4D) and 4D motion modeling are supported to analyze motion-related dose deterioration.
• The platform includes extensive data augmentation tools (inter- and intra-fraction motions) to generate training/testing data for AI and method development.

Figure 2: The tumor motion is schematically represented by the hysteresis formed by the tumor position (gray circles) at each phase of the 4DCT (a). Deformation fields (blue vectors) are generated using registration between the first phase (P1) and all others phases (b). Then, all phases can be defo