Domain-specific Cycle-GAN Augmentation

Written by Tribun on

Domain-specific Cycle-GAN Augmentation

Share this article

subscribe to our blog

[FEBRUARY 23TH, 2022]

Authors: Rutger H.J. Fick, Alizera Moshayedi, Gauthier Roy, Jules Dedieu, Stéphanie Petit, and Saima Ben Hadj - Tribun Health, Paris, France

Abstract. As the third-place winning method for the MIDOG mitosis detection challenge, we created a cascade algorithm consisting of a Mask-RCNN detector, followed by a classification ensemble consisting of ResNet50 and DenseNet201 to refine detected mitotic candidates. The MIDOG training data consists of 200 frames originating from four scanners, three of which are annotated for mitotic instances with centroid annotations. Our main algorithmic choices are as follows: first, to enhance the generalizability of our detector and classification networks, we use a state-of-the-art Residual Cycle-GAN to transform each scanner domain to every other scanner domain. During training, we then randomly load, for each image, one of the four domains. In this way, our networks can learn from the fourth non-annotated scanner domain even if we don’t have annotations for it. Second, for training the detector network, rather than using centroid-based fixed-size bounding boxes, we create mitosisspecific bounding boxes. We do this by manually annotating a small selection of mitoses, training a Mask-RCNN on this small dataset, and applying it to the rest of the data to obtain full annotations. We trained the follow-up classification ensemble using only the challenge-provided positive and hard-negative examples. On the preliminary and final test set, the algorithm scores an F1 score of 0.7578 and 0.7361, respectively, putting us as the preliminary second-place and final third-place team on the leaderboard.

Download the full article.