AI for the fully automated radiotherapy treatment chain
Chairs
Stine Korreman (AUH, Aarhus, DK), Tomas Janssen (NKI, Amsterdam, NL), Charlotte Brouwer (UMCG, Groningen, NL), Jean-Emmanuel Bibault (Universite de Paris, FR), Mark Gooding (Inpictura Ltd, UK)
Motivation
Recent years have seen the advance of automation and artificial intelligence (AI) applications within the field of radiotherapy. While initially these applications were primarily used in research settings and through in-house developments, there are now also numerous commercial options available for various applications such as segmentation, treatment planning, and adaptive radiotherapy. The incorporation of automation and AI into clinical practice in radiotherapy, a topic discussed at the ESTRO physics workshop in 2019, is currently a growing trend.
With this development, state-of-the-art applications of AI will likely transition from stand-alone components for specific purposes, to a set of tools forming a fully integrated part of daily radiotherapy treatment practice. The radiotherapy community must therefore prepare for the broader implementation of automation and AI in the full radiotherapy treatment chain including ongoing quality assurance programs.
In this workshop we will cover two aspects of the broad implementation of AI:
First, we intend to gauge the extent to which institutes currently are able to perform a fully automated treatment, from CT to irradiation. This will be achieved by means of a challenge (participation optional), running both before and during the workshop.
Second, we aim to discuss a QA program for routine use of AI in the clinic. Due to the novelty of the field, there are currently no standard approaches and reports of best practices are scarce. The program will encompass: patient specific QA (how to QA an individual model prediction); routine QA (how to QA a model after upgrades or maintenance in the RT workflow) and monitoring (how to recognize outliers or time trends).
Outcomes
The intended outcomes of the workshop are:
- A paper on state-of-the-art of fully automated RT, including results from the challenge
- A white paper on an appropriate QA program for adoption of AI in clinical practice, including best practices, challenges, and missing features
- Enhancement of the networking efforts of the ESTRO Medical Physics Future working group “Artificial Intelligence in Radiation Oncology”
- Recruitment of new members
- Development of future focus points for the working group
Invited speakers
For the pre-meeting: TBC
For the onsite workshop:
Ana Maria Barragan Montero (UCLouvain, BE)
Coen Hurkmans (Eindhoven, NL, chair of ESTRO-ACROP guideline group for development, clinical validation and reporting of AI in RT)
Participating vendors
RaySearch, MIM, Limbus, Varian, Elekta, MVision, Therapanacea
Pre-workshop programme
Pre-meeting (tentative date June 5)
An online pre-meeting will be hosted to launch the (optional) fully automated treatment planning challenge. This will consist of an invited talk on the topic of automation in radiotherapy, followed by an introduction and overview of the challenge (see below) including an opportunity for Q&A. Attendance of this pre-meeting is not required either to participate in the challenge or in the workshop.
Online challenge (May-Sept)
An (optional) online challenge will be launched to address the question "How close are we to fully automated radiotherapy treatment planning?" The challenge will be open May-Sept.
More details can be found at https://auto-rtp.grand-challenge.org/auto-rtp/
Onsite workshop programme (13-14 2023)
|
Day 1 - October 13
|
|
09:00
|
10:00
|
|
Plenary session, invited speaker
|
|
10:00
|
10:30
|
|
Coffee break
|
|
10:30
|
12:30
|
|
|
|
10:30
|
10:45
|
|
Introduction to workshop
|
|
10:45
|
11:15
|
|
Invited lecture Coen Hurkmans, tentative title: "The ESTRO-ACROP guideline for development, clinical validation and reporting of AI in RT"
|
|
11:15
|
11:30
|
|
Mark Gooding and Jean-Emmanuel Bibault: Overview of challenge results "Fully automated treatment planning"
|
|
11:30
|
12:00
|
|
Selected participants contributions from challenge
|
|
12:00
|
12:30
|
|
Discussion (30 minutes)
|
|
12:30
|
13:30
|
|
Lunch
|
|
13:30
|
15:30
|
|
|
|
13:30
|
14:00
|
|
Invited lecture Ana Barragan Montero, tentative title: "Application of machine learning in the RT workflow: risks and QA options"
|
|
14:00
|
14:30
|
|
Selected participants contributions (focus on QA)
|
|
14:30
|
15:30
|
|
Discussion (60 minutes)
|
|
15:30
|
16:00
|
|
Coffee break
|
|
16:00
|
17:00
|
|
Plenary session progress reports from groups
|
|
Day 2 - October 14
|
|
08:00
|
10:00
|
|
Exercises in groups
|
|
|
|
|
Group A: Onsite challenge
|
Group B: QA risk analysis
|
|
10:00
|
10:30
|
|
Coffee break
|
|
10:30
|
12:30
|
|
|
|
10:30
|
11:00
|
|
Wrap up of previous discussions and exercises
|
|
11:00
|
11:45
|
|
Brainstorm on future steps and division into groups
|
|
11:45
|
12:30
|
|
Group work on future steps
|
|
12:30
|
13:30
|
|
Lunch
|
|
13:30
|
14:30
|
|
|
|
13:30
|
14:00
|
|
Group work on future steps (continued)
|
|
14:00
|
14:30
|
|
Wrap up of workshop
|
|
14:30
|
15:45
|
|
Plenary wrap up - reporting from all groups and closing
|