Open Ground Truth Training Network – Simulated MRI for medical image analysis research
Five abstracts have been accepted for poster presentation at the 28th Annual Meeting of the International Society of Magnetic Resonance in Medicine (ISMRM). The meeting was supposed to take place in May 2020 in Sydney, Australia, but was rescheduled to 8-13 August in Paris, France. The abstract titles can be found here.
4D Semantic Cardiac Magnetic Resonance Image Synthesis on XCAT Anatomical Model
Available on Openreview and arXiv.
We are thrilled to share with you our recent paper accepted for @midl_conference 2020 presenting a novel hybrid controllable image generation method for synthesizing anatomically meaningful 3D+t labelled CMR images. Check out how we efficiently combine a physics-driven anatomical model with a SOTA data-driven generative adversarial network to generate realistic-looking cardiac MR images. As a part of the OpenGTN research, we step towards tackling the limited availability of medical data for deep learning training.
With much regret we have to inform you that we have to cancel the openGTN Workshop planned for 27 March 2020. Especially in the province of Noord-Brabant, in which Eindhoven is located, we have a significant outbreak of the Corona virus. As a consequence, Dutch government has recommended to work as much as possible at home and avoid unnecessary contacts. We hope we can reschedule the workshop before Summer 2020, and we hope that you will then still be interested to attend.
The openGTN public workshop on “Realism in MRI simulation for deep learning training” attracts a lot of attention. There are only 10 free seats left! The registration is soon expected to close. More information and a registration form can be found here.
The openGTN project organizes its first public workshop on the theme “Realistic magnetic resonance image simulation” on 27 March 2020 at the Eindhoven University of Technology, The Netherlands.
The three openGTN early-stage researchers will present the approaches investigated so far to generate realistic simulated MRI data and will discuss the use of these data to improve deep-learning based MRI segmentation. Thereafter three renowned, invited speakers (prof. Stoecker, prof. Rueckert and prof. Kozerke) will give their view on the subject, from different perspectives. The workshop will close with a discussion about how to make MRI simulations even more realistic.
Workshop attendance is free of charge, but registration is required (note: limited seating available!). More information and a registration form can be found here.
The first openGTN project results will be presented at the ISMRM Benelux 2020 scientific conference, 24 January 2020, Arnhem, The Netherlands.
Three presentations will be given:
“Improving the generalization capability of deep learning-based algorithms for ventricular cavity segmentation using simulated cardiac MR images”
(oral presentation by Yasmina Al Khalil)
“Towards generating realistic and heterogeneous cardiac MR simulated image database for deep learning based image segmentation algorithms”
(poster presentation by Sina Amirrajab)
“Pipeline for Simulating Realistic Anatomically Variable Normal Young, Aging and Diseased Brain MRI”
(poster presentation by Aymen Ayaz)
The openGTN project can now also be approached via Facebook and Twitter, see the links in the footer of this website.
There is now an openGTN discussion group on LinkedIn:
You can request to become member via our contact form
Updated CVs of the involved Early Stage Researchers Aymen Ayaz, Yasmina Al Khalil and Sina Amirrajab have been added to the team page
We have updated our privacy policy, which you can read here.