Automated Radiation Therapy Treatment Planning for Volumetric Arc Therapy in Prostate Cancer
Sandeepan Ganguly1, Jingqiao Zhang2, Jason Lima1, Yijian Cao1,2, Jameson Baker1,2, Jenghwa Chang1,2
1Medical Physics, Hofstra University, New York; 2Radiation Medicine, Northwell Health, New York
Innovation/Impact:
Our study provides a robust and quick software tool for manipulating HID devices irrespective of TPS during the clinical workflow. By automating the workflow and machine learning inference, the operation time can be significantly reduced as well as the human errors and stress. The software interface can be extensively implemented on the different types of planning procedures.
PURPOSE / OBJECTIVES
To develop a robust and interoperable software interface using scripting and optical character recognition (OCR) for assisting treatment planning procedures on a commercial treatment planning system (TPS). Treatment Planning is a very involved tasks with a lot of variable that requires specialized skills and experience.
MATERIAL & METHODS
The general software interface was developed for human interface devices (HIDs) by using AutoIt scripting language and OCR capabilities. The upgraded AUTOFLOW functions according to the action table, which was preset according to our planning protocol template. This new software was tested on the Eclipse TPS in our clinics for three (contouring, external beam planning and optimization) automatic planning functions. Based on our clinical planning protocol, the planning starts from post contour approval, AUTOFLOW guides the Eclipse to create auxiliary structures, set up planning fields then initiate plan optimization. During the optimization, the AUTOFLOW fills the constraint parameters and monitors the objective goals achievement by using OCR. Finally, the planning parameters are updated with our developing parameter prediction model.
RESULTS
The new general interface AUTOFLOW was used to automatically assist the VMAT optimization for five prostate cases. 28 patients’ optimization objectives were used to create the statistical modeling and the optimized initial parameters of objectives were provided. More than 113 human HID operations on computers were removed from a single plan procedure. The interface reduced the VMAT planning time considerably and still achieved the constraint goals. This saves clinicians time waiting for the computation to complete. We have developed an IMRT QA System that can generate verification plans and perform patient specific QA with minimal human intervention. This research hence shows the potential of Automation and provides a general framework to achieve fully autonomous adaptive treatment planning system.
SUMMARY / CONCLUSION
The developed software interface can be used to perform VMAT planning in a commercial clinical Treatment Planning System and fully customizable to work around individual clinic’s workflow It can automatically assist planner operation and provide programming interface to 3rd party software packages. The future works will focus on improving the prediction of optimization parameters and testing it on other treatment sites
For more information, please contact:
• Sandeepan Ganguly sganguly1@pride hofstra edu
• Jenghwa Chang, PhD Jenghwa chang@hofstra edu
Key Results: The study builds up a software interface for the workflow of auto treatment planning. The whole system layout is shown in Figure 1. AutoFlow interface attaches the Eclipse Planning System. The green arrows indicate the treatment planning workflow. Human being still can access the workflow as needed. Figure 2. The Autoflow interface with respective dropdown list to control 3 interfaces in Eclipse during the planning.
