Domain Generalisation for Mitosis Detection Exploting Preprocessing Homogenizers.

Published in ESMO MAP, 2021

The detection of mitotic figures in histological tumor images plays a vital role in the decision-making of the appropriate therapy. However, tissue preparation and image acquisition methods degrade the performances of the deep learning-based approaches for mitotic figures detection. MIDOG challenge addresses the domain-shift problem of this detection task. In an endeavour to reduce this domain shift, we propose a pre-processing autoencoder that is trained adversarially to the sources of domain variations. The output of this autoencoder, exhibiting a uni- form domain appearance, is finally given as input to the retina-net based mitosis detection module.

Citation

‘Nasser, S., Kurian NC, Sethi A. Domain Generalisation for Mitosis Detection Exploting Preprocessing Homogenizers. International Conference on Medical Image Computing and Computer-Assisted Intervention 2021 Sep 27 (pp. 77-80). Springer, Cham.’

Paper Link