A new color augmentation method for deep learning segmentation of histological images

Yang Xiao (1) , Etienne Decencière (1) , Santiago Velasco-Forero (1) , Hélène Burdin (2) , Thomas Bornschlögl (3) , Françoise Bernerd (3) , Emilie Warrick (3) , Thérese Baldeweck (3)

1 CMM - Centre de Morphologie Mathématique
2 ADCIS - Société ADCIS
3 L'Oréal Recherche - L'Oréal Recherche France

This paper addresses the problem of labeled data insufficiency in neural network training for semantic segmentation of color-stained histological images acquired via Whole Slide Imaging. It proposes an efficient image augmentation method to alleviate the demand for a large amount of labeled data and improve the network's generalization capacity. Typical image augmentation in bioimaging involves geometric transformation. Here, we propose a new image augmentation technique by combining the structure of one image with the color appearance of another image to construct augmented images on-the-fly for each training iteration. We show that it improves performance in the segmentation of histological images of human skin, and also offers better results when combined with geometric transformation .

Mots clés: color-stained slide; deep learning; image segmentation; color transfer; histopathology; Fontana Masson; skin

Domaines: Traitement des images [eess.IV]; Vision par ordinateur et reconnaissance de formes [cs.CV]

Yang Xiao, Etienne Decencière, Santiago Velasco-Forero, Hélène Burdin, Thomas Bornschlögl, et al.. A new color augmentation method for deep learning segmentation of histological images. 2019 IEEE 16th International Symposium on Biomedical Imaging (ISBI), Apr 2019, Venise, France. ⟨hal-02167903⟩