Deep-Learning MR-only Radiation Therapy

This project seeks to improve radiation therapy for cancer patients by enhancing MR imaging so that CT scans can be eliminated from the process.

The solution realises the full potential of multi-parametric MRI and Deep Learning, to reduce time and stress for patients while providing images that can make radiation therapy more effective.



Approximately 1-in-3 people will develop cancer in their lifetime, and about half will receive radiation therapy, which uses radiation to destroy cancer cells while sparing surrounding healthy tissue as much as possible. Currently, this requires making tumour images with multiple MRI and CT scans, which is time-consuming, expensive and stressful for patients. Comprehensive MR-only Radiation Therapy will benefit both patients and healthcare professionals.


We provide a truly collaborative environment, including clinical, academic and commercial organisations comprising expertise in medical imaging, radiotherapy physics, radiation oncology, machine learning and project management. Together, we have a unique vision of how our project can benefit our patients and bring this into routine clinical use based on experience in the U.K., the Netherlands, Germany and Hungary.

The project

This project provides a solution to improve imaging for radiation therapy, making it possible to optimise treatment for each individual patient. Radiation therapy requires 3D medical images that detect borders of the tumour and organs – to guide the use of radiation so that it primarily kills cancer cells. Current imaging is based on a complicated workflow requiring multiple scans with magnetic resonance imaging (MRI) and computer tomography (CT).

The process is expensive, inconvenient for patients and prone to positioning and registration errors, and CT scans expose patients to radiation. The Deep MR-only RT project aims to develop a smooth MRI-only imaging workflow. Comprehensive multi-parametric MRI in combination with novel RF coil concepts maximises scan efficiency and image utilisation, as required for accurate therapy planning.

Deep Learning will automate and improve tumor target delineation and identification of organs at risk, eliminate potential errors, reduce overall recurrence rates and decrease unpleasant side effects. Together, these technologies provide all the imaging information needed to plan accurate and reproducible plans for radiation therapy with a single MRI examination.

The partners have already invested more than €8 million for activities related to this project. With the additional €2.6 million funding from EIT Health and co-funding of €342 000 from partners, results are anticipated in three years’ time.


Deep MR-only RT will provide a comprehensive one-stop solution with improved clinical outcome at reduced cost.

Benefits for patients include: shorter and quieter MR scanning, which improve patient comfort, and the ability to avoid the radiation of CT scans.

Benefits for doctors are elimination of CT scans, which prevents alignment errors due to repeated patient set-ups and nonrigid body motion and may also improve long-term patient survival for young and pediatric patients.

Why this is an EIT Health project

Deep MR-only RT is aligned with the basic EIT Health goal of improved medical care. It is also in keeping with the Focus Area of “Care Pathways” because it promises improved radiation therapy treatment with better outcomes, due to improved imaging.

External partner

  • Szeged University
  • King’s College London
Dr Florian Wiesinger
| Principal Scientist | GE Healthcare Munich, Germany; King’s College London, UK
Dr. Peter Bencsik
| Consortia Research Leader | GE Healthcare