Batch Export: An automated framework for curated data extraction via the Eclipse treatment planning system

批量导出:通过 Eclipse 治疗计划系统自动提取精选数据的框架

阅读:1

Abstract

BACKGROUND: Deep-learning models are useful for radiation therapy tasks such as 3D dose prediction, auto-contouring and auto-planning. These models require large training datasets to obtain clinically acceptable results. Currently, the process of exporting DICOM-RT data from the Eclipse treatment planning system (TPS) is often tedious and becomes unscalable when approaching the magnitude of hundreds of patients. Thus, an efficient procedure to obtain this data would be effective for downstream research applications. PURPOSE: The purpose of this study was to simplify and improve the efficiency of data retrieval from the Eclipse TPS. To do so, we have created an application which exports patient treatment plans and associated images and structure sets in a parallel, streamlined process. METHODS: The application was made using C# .NET and the Prism library to create a graphical user interface (GUI). EvilDICOM was integrated within the GUI to facilitate the connection to the Eclipse patient database and obtain patient plans. Data export using our application was compared to manual export using the Eclipse Export module; specifically, timing data as a function of the number of Digital Imaging and Communications in Medicine (DICOM) files exported was assessed. RESULTS: Utilizing the application was faster than manually exporting via the TPS in cases with more than one patient. When attempting to perform an export of 20 patients' treatment data (∼3 000 DICOM files, including the plan, structure set, dose, and all slices of the CT image), our application took 10.22 min while manual export took 22.93 min. Our application proved to be a linear-time process and scalable to over, but not limited to, 17 000 DICOM files. CONCLUSIONS: We have developed an open-source application to rapidly obtain patient data from Eclipse in a scalable manner. This tool addresses the challenges of manually exporting DICOM files in large magnitudes and increases the feasibility for processes like machine learning model training.

特别声明

1、本页面内容包含部分的内容是基于公开信息的合理引用;引用内容仅为补充信息,不代表本站立场。

2、若认为本页面引用内容涉及侵权,请及时与本站联系,我们将第一时间处理。

3、其他媒体/个人如需使用本页面原创内容,需注明“来源:[生知库]”并获得授权;使用引用内容的,需自行联系原作者获得许可。

4、投稿及合作请联系:info@biocloudy.com。