The Radiation Planning Assistant is being developed to offer a suite of fully automated contouring and radiotherapy planning tools that we aim to eventually make available through a web-based interface. Note, the RPA is in its Research Use Only phase, and will not be available for clinical use until it receives its medical device clearance from the US Food and Drug Administration.
The process starts with the user uploading and accepting a CT scan, and completion of a Service Request form. The planning process then starts automatically. First the patient data is automatically transferred to a secure location, where it is de-identified before being sent to the RPA servers.
For complex plans (like VMAT plans for head and neck cancer treatments), the system first performs automatic contouring, which the user then reviews and edits in their own treatment planning system before upload to the RPA system for final plan creation. For simple plans (like 4-field box plans for cervix treatments), the plan is automatically completed without additional user interaction.
In all cases the final plan must be downloaded to the user’s own treatment planning system where the dose is recalculated and the user makes any necessary edits before the plan can be used clinically. The RPA should be used as a tool to improve workflows – final evaluation of the plan, including evaluation of its suitability for clinical use and assessment of all plan parameters, should be performed as part of the routine quality assurance process at the user institution. Additional information on the different services offered by the RPA can be found under Services.
The RPA is powered by artificial intelligence tools to perform preparation and contouring tasks. Plan optimization and dose calculation is automatically performed in the Eclipse Treatment Planning System (Varian Medical Systems), with dose verification performed using Mobius (Varian Medical Systems). The RPA uses purposely developed algorithms to perform a series of quality assurance checks, designed to flag suspicious contours or plans to the user for further investigation. Further details of the algorithms can be found in our publications. Final contours and plans should be imported into the user's own treatment planning system for review, edit, and re-calculation of dose.
We are investigating how to make the RPA Program financially self-supporting, without making it unaffordable to clinics in low- and middle-income countries. Our goal is to make the RPA available to these clinics at minimal or zero cost.
Collaboration & Funding
Our main international collaborators during the development of the RPA have been Tygerberg Hospital (Stellenbosch University) and Groote Schuur Hospital (University of Cape Town), both in South Africa, and Stanford University (USA). We also have collaborations with the University of the Free State (South Africa), Benavides Cancer Institute (Santo Tomas University, Philippines), Guy’s & St. Thomas’ Hospital (United Kingdom) and Gaborone Private Hospital (Botswana). We are constantly working to expand this network, including our Latin American RPA Consortium, focusing on the needs of Latin American countries.
The RPA project was started with funding from the National Cancer Institute (co-PIs: Laurence Court, Beth Beadle), and has since received additional development funding from Varian Medical Systems (PI: Laurence Court), the Wellcome Trust (PI: Laurence Court), and the University of Texas MD Anderson Cancer Center (PI: Laurence Court).
The registration process is currently only open for testing to select collaborators. In order to ensure that clinics are able to use the RPA in their own clinical workflows, the registration process includes independent audits of the radiation outputs (all photon beams), and an independent audit of the local treatment planning system. Currently these audits are offered free to our collaborators.