OncoReg - Challenge evaluation for the integration of expert anatomical knowledge from retrospective data in oncological computed tomographic image registration with artificial intelligence (BMBF)

More than 1.7 million people die from lung cancer each year, making it one of the deadliest oncological diseases worldwide. Image registration and machine learning are fueling the development of novel medical technologies that have the potential to significantly change the diagnosis and treatment of lung cancer. At the same time, modern cancer research is accumulating increasingly large data sets that can be leveraged using advanced artificial intelligence methods.

The project OncoReg aims at the preparation and implementation of a data challenge for the registration of oncological image data of the lung. It aims to provide low-threshold access to high-quality data from translational, biomedical cancer research and routine oncology care, and to promote a culture of data sharing for research purposes. The planned Challenge aims to develop 3D registration methods for computed tomography image data from radiation therapy to address clinically relevant questions in oncology.For example, information from high-resolution fan-beam computed tomography images acquired for therapy planning in radiotherapy will be transferred to low-resolution interventional cone-beam CT images using novel 3D registration techniques.
The data foundation of the Challenge is a curated dataset based on supplementary manual annotations by radiologists on already publicly available image material, as well as a previously unpublished data from the project partner UKSH. By providing high-quality annotations of anatomical structures, landmarks and lesions, submissions of innovative algorithms and AI models specifically targeting the follow-up of lung lesions and 4D radiotherapy of oncological lung diseases are expected.





The scientific objectives of the project are:
- the curation of an annotated multimodal fan-beam/cone-beam CT dataset with manual segmentations of various structures, including tumor, organs at risk and other anatomies important for radiotherapy planning, as well as an additional annotated dataset for follow-up,
- the development and provision of deep learning segmentation models and AI-based image registration modules, and
- the design and implementation of a platform for privacy-preserving algorithm training.

OncoReg goes beyond solving a single task within oncological image analysis, but also aims to provide best-practice solutions for the research community - adaptable to new applications.

In addition to the University of Lübeck, UKSH Lübeck is involved.

BMBF Project Funding (2022-2024) 251.665€ (IMI 125.357€)


Selected Publications:

Hering, A., Hansen, L., Mok, T. C., Chung, A. C., Siebert, H., Häger, S., ... & Heinrich, M. P. (2022). Learn2Reg: comprehensive multi-task medical image registration challenge, dataset and evaluation in the era of deep learning. IEEE Transactions on Medical Imaging, 42(3), 697-712.