MItosis DOmain Generalization Challenge 2022

Latest news ^3:

The preprint of the MIDOG 2022 challenge report is now available here.

Latest news^2:

Unfortunately, we've had an error in the previous calculation of the additional Average Precision (AP) metric. Please find out all details in this blog post.

Latest news: 

We are excited to announce that the MIDOG++ dataset, an extension of the MIDOG2022 training data set has just been released. We have published a data descriptor paper for a full description of the data set. To get a glimpse, we recommend reading the blog post about it

• Mitosis detection is a key component of tumor prognostication for various tumors.
• Modern deep learning architectures provide detection accuracies for mitosis that are on the level of human experts.
• Mitosis is known to be relevant for many tumor types, yet, when trained on one tumor / tissue type, the performance will typically drop significantly on another.

1. Detect mitotic figures (cells undergoing cell division) from histopathology images (object detection). 
2. You will be provided with images from 6 different tumor types, 5 out of which are labeled. In total the set consists of 405 cases and includes 9501 mitotic figure annotations in the training set.
3. Evaluation will be done on ten different tumor types with the F1 score as main metric.

• Register on the challenge website and download the data set.
• Select a track (see below) that you want to participate in.
• Submit docker container(s) with your algorithm(s) (template provided here: https://github.com/DeepPathology/MIDOG_reference_docker).
• Provide a short paper (preprint) about your method. A template can be found under this link.  
• Note: Only your own work will qualify as a submission.

• There's different ways to approach this task, thus we split up the challenge into two tracks that participants can participate in (either in one or in both).
• Track 1 prohibits the use of any additional (i.e. not provided by the challenge) data and also any kind of manual or semi-automatic generation of any kind of label information.
• In Task 2, use of any publicly available data sets and additional labels is permitted. If you want to make use of previously private labels, please have a look at the related blog post.