Next generation MR-guided radiotherapy: AI applications for planning and image guidance

 

Chairs: Davide Cusumano (Italy), Marco Fusella (Italy), Lorenzo Placidi (Italy)

 

Motivation

Magnetic Resonance guided Radiotherapy (MRgRT) is a novel therapeutic approach that is emerging as a promising reality in the clinical radiotherapy practice.

The use of an on-board MR scanner allows clearer soft-tissue visualization, thus making possible the reduction of the therapy volumes and allowing effective compensation of the inter- and intra-fraction motion variability, through dedicated strategies such as online adaptive RT procedures (ART) and online MR image acquisition.

Being a discipline in its infancy, MRgRT offers different opportunities of physical and clinical innovations. Multiple research groups have recently proposed different innovations to improve the clinical MRgRT practice, with the aim to automate the manual procedures, making the clinical workflow more reliable and less operator dependent.

One of the most interesting perspectives in MRgRT is represented by the integration of artificial intelligence (AI) systems. AI can support the entire clinical team (radiation oncologists, medical physicist and radiation therapists) in the daily workflow, allowing the standardisation of the treatment plan quality, increasing the safety of the process (through standardization and automation of processes such as total dose accumulation and QA) and reducing the overall treatment time, which represents a key-point during the online ART. 

The most interesting AI applications in MRgRT are related to the generation of synthetic Computed Tomography (sCT) images from MRI and the automation of manual procedures, such as segmentation and treatment planning. With the advent of the integration of multiple AI approaches into MRgRT clinical practice, it assumes a central role the need of dedicated quality assurance procedures for an optimised and appropriate clinical implementation. Thanks to the support of AI, the next generation of MR-Linac will pave the way towards the MRI-only clinical workflows and online dose painting approaches based on modern MR-based biomarkers.

 

The present workshop will be focused on the main topics of the next generation MRgRT, concerning the development, validation, implementation, commissioning and quality assurance of AI systems for an appropriate and optimized clinical practice. The aim is to create a network of physicists who want to share ideas and experiences, in the perspective of contributing to the improvement of this therapeutic approach.

 

Outcome

The potential outcome of the workshop is to exchange clinical and physical experiences, sharing criticalities related to the AI applications for planning and image guidance of MRgRT, both for low field and high field technologies. Starting from the detailed analysis of the present challenges, novel approaches will be discussed. Potential outcomes of the workshop will include:

  • Establishing dedicated working groups to discuss the application of AI (synthetic CT generation, automatic contouring, automatic planning and online MR imaging) in MRgRT system;
  • Providing recommendations and guidance documents;
  • Survey among centres about next generation MR-Linac wish list and implementation, also discussing with the MRgRT vendors;
  • White paper: status and goals for the future;
  • Assembling an international expert group to facilitate the widespread clinical implementation for the next generation of MRgRT, promoting protocols of privacy-preserving data sharing and supporting the diffusion of next generation MRgRT.

 

Invited speakers

Tomas Janssen (NKI Amsterdam) – MRgRT: state of the art and physical perspectives

Jennifer Dhont (Institut Jules Bordet - Bruxelles) – AI: the basis and the principles

Chiara De-Colle (Uni Tubingen) - Challenges of AI in MRgRT: the clinical perspective

Stefanie Corradini (LMU Munich) - Challenges of AI in MRgRT: the clinical perspective

 

Vendor contacts

  • Viewray, Elekta, Magnettx (Gino Fallone, https://www.magnettx.com), Australian MR-LINAC project (lois.holloway@health.nsw.gov.au)
  • Therapanacea, LimbusAI, Mirada, Mim, Spectronic, Philips (MRCAT), Siemens, MVision, RayStation

 

Programme

ONLINE 

6 September 

14:00-14:15 

Welcome, Workshop introduction and programme overview 

14:15-14:40 

Presentation of Topic 1: AI for MRI-only workflow for Radiotherapy treatments (sCT, planning, QA) 

14:40-15:05 

Presentation of Topic 2: AI for image guidance, adaptive (real-time autocontouring and planning and AI-QA) 

15:05-15:30 

Presentation of Topic 3: AI for patient outcome: MRI-based biomarkers (radiomic, quantitative imaging) 

16:00-16:30 

Presentation from participants groups (5 min each) - Q/A – Discussion 

16:30-17:00 

Software presentation for discussion and topic selection (MIRO)

 

Day 1

Friday 7 October

09:00-09:15

Introduction of the meeting (ESTRO)

09:15-10:00

Opening lecture – XY (ESTRO)

10:00-10:30

Coffee break

10:30-11:00

Introduction of the meeting – Overall Chair of workshop (all)

Program and WG presentations.

11:00-11:30

Invited speaker 1 – AI the basis and the principles – J. Dhont (all)

30 min presentation + 15 min discussion

11:30-12:00

Invited speaker 2 – Challenges of AI in MRgRT – the physical perspective T. Janssen (all)

35 min presentation + 15 min discussion

12:00-12:30

Next-generation MR-Linac: what does mean for you?

Elekta, ViewRay, Magnettx, Australian MR-LINAC

12:30-13:00

Discussion

13:00-14:00

Lunch

14:00-14:30

WG presentation (literature review and topics definition in WG) – WG Chair

(WG activity)

14:30-15:30

Single presentations of participants and discussion (max 10 min each, WG activity)

15:30-16:00

Coffee break

16:00-17:00

Discussion on WG topic and definition of possible workshop outcomes (WG activity)

17:00-18:00

Chair wrap-up and discussion (only Chairs)

 

 

Day 2

Saturday 8 October

08:30-08:45

Wrap-up: brief overview of day 1 (all)

08:45-09:45

Invited speaker 3 and 4 – Challenges of AI in MRgRT: the clinical perspective – S. Corradini and C. De-Colle (all)

50 min presentation + 10 min discussion

09:45-11:00

Definition of outcome-roadmap, consortium formation, methodology paper (WG activity)

11:00-11:30

Coffee break

11:30-12:00

Presentation of activities and planned outcomes of WG1 (all)

15 min presentation of WG members + 15 min interactive discussion with all participants

12:00-12:30

Presentation of activities and planned outcomes of WG2 (all)

15 min presentation of WG members + 15 min interactive discussion with all participants

12:30-13:00

Presentation of activities and planned outcomes of WG3 (all)

15 min presentation of WG members + 15 min interactive discussion with all participants

13:00-13:30

Wrap up of the different topic workshops (all)

10 min presentation per Chair/responsible of WG

13:30-14:30

Lunch

14:30-15:30

Wrap up of the different topic workshops (ESTRO)

15:30-15:45

Closure

Legend

ESTRO: All participants of ESTRO Physics workshop

All: All participants of Next generation MR-guided radiotherapy: AI application for planning and image guidance

WG: Participants of each working group