Smart Learning of Click and Refine for Nuclei Segmentation on Histology Images

Antoine Habis (1, 2, 3) , Vannary Meas-Yedid (1) , Daniel Felipe Gonzalez Obando (1) , Jean-Christophe Olivo-Marin (1) , Elsa Angelini (2, 3, 4)

1 BIA - Analyse d'images biologiques - Biological Image Analysis
2 LTCI - Laboratoire Traitement et Communication de l'Information
3 IP Paris - Institut Polytechnique de Paris
4 Imperial College London

Deep learning has proven to be a very efficient tool to help pathologists analyze Whole Slide Images (WSI) toward automated classification or segmentation of detailed structures such as nuclei, glands or glomeruli. These objects are particularly relevant for disease diagnosis and staging. Many deep learning methods have shown impressive performance but are still imperfect, while manual segmentation has poor inter-rater agreement. In this paper, we propose a patch-level automated correction of a given baseline initial segmentation, based on deep-learning of segmentation errors and downstream local refinements. Results on the MoNuSeg and PanNuke test datasets show significant improvement of nuclei segmentation quality.

Keywords: Deep Learning; image segmentation; Histopathology

Domains: Image Processing [eess.IV]; Artificial Intelligence [cs.AI]; Computer Vision and Pattern Recognition [cs.CV]; Machine Learning [cs.LG]

Antoine Habis, Vannary Meas-Yedid, Daniel Felipe Gonzalez Obando, Jean-Christophe Olivo-Marin, Elsa Angelini. Smart Learning of Click and Refine for Nuclei Segmentation on Histology Images. 2022 IEEE International Conference on Image Processing (ICIP), Oct 2022, Bordeaux, France. pp.2281-2285, ⟨10.1109/ICIP46576.2022.9897496⟩. ⟨pasteur-03927669⟩