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

Area of Science:

• Medical Physics
• Radiotherapy
• Machine Learning

Background:

• Deep learning models require large datasets for radiation therapy tasks like dose prediction.
• Current DICOM-RT data export from Eclipse TPS is inefficient and unscalable for large datasets.
• Efficient data retrieval is crucial for advancing downstream research applications.

Purpose of the Study:

• To simplify and enhance the efficiency of patient data retrieval from the Eclipse TPS.
• To develop a streamlined application for parallel export of treatment plans, images, and structure sets.
• To overcome the limitations of manual data export for large-scale research.

Main Methods:

• Developed a C#.NET application with a GUI using the Prism library.
• Integrated EvilDICOM for seamless connection to the Eclipse patient database.
• Compared application data export times against manual export methods based on DICOM file volume.

Main Results:

• The application significantly reduced export times compared to manual methods, especially for multiple patients.
• Exporting 20 patients' data (∼3000 DICOM files) took 10.22 minutes with the application versus 22.93 minutes manually.
• The application demonstrated linear-time performance and scalability for over 17,000 DICOM files.

Conclusions:

• An open-source application was created for rapid and scalable patient data acquisition from Eclipse TPS.
• The tool effectively addresses challenges associated with manual DICOM file export in large volumes.
• This facilitates machine learning model training and other research requiring extensive patient data.