Domain Adversarial RetinaNet as a Reference Algorithm for the MItosis DOmain Generalization Challenge

Abstract

Assessing the Mitotic Count has a known high degree of intra- and inter-rater variability. Computer-aided systems have proven to decrease this variability and reduce labeling time. These systems, however, are generally highly dependent on their training domain and show poor applicability to unseen domains. In histopathology, these domain shifts can result from various sources, including different slide scanning systems used to digitize histologic samples. The MItosis DOmain Generalization challenge focused on this specific domain shift for the task of mitotic figure detection. This work presents a mitotic figure detection algorithm developed as a baseline for the challenge, based on domain adversarial training. On the challenge's test set, the algorithm scored an F${}_1$ score of 0.7183. The corresponding network weights and code for implementing the network are made publicly available.

arXiv commentThis is the long version of the original pre-print. Due to a bug in our automatic threshold computation the detection threshold of our model changed from 0.62 to 0.64. This value was not optimized on any other images but the validation split of the MIDOG training set. 9 pages, 4 figures, 1 table

References in corpus (4)

• UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction
• Unsupervised Domain Adaptation by Backpropagation
• Are pathologist-defined labels reproducible? Comparison of the TUPAC16 mitotic figure dataset with an alternative set of labels
• Quantifying the Scanner-Induced Domain Gap in Mitosis Detection