Join


OncoReg: Medical Image Registration for Oncological Challenges 


The OncoReg challenge finished with our virtual workshop on February 20th, 2024:

  • 2:30 PM - 2:50 PM:      Introduction & Presentation of the Challenge  
  • 2:50 PM - 3:20 PM:      Invited Speaker: Dr. Malte Sieren “Shaping the Diagnostic Landscape: AI's Role in the                                                       Future of Clinical Radiology"
  • 3:20 PM  - 4:35 PM:     Presentation and Discussion of three Methods
  • 4:35 PM - 5:00 PM:      Presentation of Results and Awards


Congratulations!  

If you are interested in evaluating your method on our hidden data, you are welcome to send us a message. We offer a continuation of the evaluation beyond the completion of the challenge. 

For further information on the data and on type 3 challenge submissions you can find our Kick-Off slides here. Please also see our latest forum post and our OncoReg GitHub for submission details.  


About OncoReg

The Learn2Reg challenge series has been established over recent years to develop and benchmark (primarily learning-based) medical image registration algorithms for a variety of tasks. Within our new spin-off OncoReg our aim is to shift the focus to new datasets and evaluations that address real-world oncological challenges. 

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 data challenge is intended to contribute to better research into oncological diseases of the lung and at the same time promote data sharing for research purposes. The competition is multidisciplinary: expert anatomical knowledge meets artificial intelligence methods. AI models are designed to help monitor the progression of lung cancer during treatment and improve radiation therapy. For the competition, radiation therapists will mark important structures in CT images of lung tumor patients including the tumor, organs at risk and anatomical landmarks.


Fig. 1: High-resolution fan-beam computed tomography image of a patient with non-small cell lung cancer from the ThoraxCBCT dataset with manual annotations. Left: Bone structures and organs at risk in the thorax. Middle: Tumor. Right: Lung and trachea.


First, we released a challenge task for registering diagnostic high-quality CT and challenging low-dose intra-operative cone-beam CT (ThoraxCBCT). Second, we extended the current NLST task from Learn2Reg on intra-patient follow-up for lung-imaging by providing lesion annotations and evaluation. The results of these tasks as type 1/2 challenge were discussed at our MICCAI workshop. 

Following our Kick-Off after MICCAI  the submission of new entries of methods for a type 3 challenge were accepted. The methods were applied or retrained on a private dataset from clinical routine (radiotherapy at university hospital Lübeck) with manual expert annotations for evaluation.

The results for this out-of-domain evaluation, which decouples model development from the actual data were analysed and discussed at a virtual winter workshop in February 2024 with free participation for all participants or interested parties. 



We highly encourage prospective participants to publish open-access journal papers in our special issue on image registration in the MELBA journal (deadline November 30): https://www.melba-journal.org/blog/012-special-issue-image-registration.html.  

However, this is not a prerequisite and we are welcoming both published and new methods with or without learning.

We invite interested researchers from academia and industry to apply either learning based or conventional medical registration algorithms to the new oncological tasks. 


Important Dates: 

  • After MICCAI 2023: Participants algorithm development 
  • December 6, 2023: KickOff and virtual help session for type 3 challenge submission
  • January 31, 2024: Deadline for submitting final dockers 
  • February 20, 2024, 2:30 - 5 pm CET: Virtual winter workshop presenting results

The organisers are  Mattias Heinrich, Veronika Zimmer, Wiebke Heyer, Li Zi and Alessa Hering.

Please contact us at learn2reg@gmail.com for any queries.