Topology-aware method to segment 3D plan tissue images

From LRDE

Revision as of 07:24, 27 October 2022 by Bot (talk | contribs) (Created page with "{{Publication | published = true | date = 2022-10-25 | authors = Minh Ôn Vũ Ngoc, Nicolas Boutry, Jonathan Fabrizio | title = Topology-aware method to segment 3D plan tissue...")
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

Abstract

The study of genetic and molecular mechanisms underlying tissue morphogenesis has received a lot of attention in biology. Especially, accurate segmentation of tissues into individual cells plays an important role for quantitative analyzing the development of the growing organs. Howeverinstance cell segmentation is still a challenging task due to the quality of the image and the fine-scale structure. Any small leakage in the boundary prediction can merge different cells together, thereby damaging the global structure of the image. In this paper, we propose an end-to-end topology-aware 3D segmentation method for plant tissues. The strength of the method is that it takes care of the 3D topology of segmented structures. The keystone is a training phase and a new topology-aware loss - the CavityLoss - that are able to help the network to focus on the topological errors to fix them during the learning phase. The evaluation of our method on both fixed and live plant organ datasets shows that our method outperforms state-of-the-art methods (and contrary to state-of-the-art methods, does not require any post-processing stage). The code of CavityLoss is freely available at https://github.com/onvungocminh/CavityLoss

Documents

Bibtex (lrde.bib)

@InProceedings{	  movn.22.nips,
  author	= {Minh \^On V\~{u} Ng\d{o}c and Nicolas Boutry and Jonathan
		  Fabrizio},
  title		= {Topology-aware method to segment 3D plan tissue images},
  booktitle	= {Thirty-sixth Conference on Neural Information Processing
		  Systems},
  year		= 2022,
  abstract	= {The study of genetic and molecular mechanisms underlying
		  tissue morphogenesis has received a lot of attention in
		  biology. Especially, accurate segmentation of tissues into
		  individual cells plays an important role for quantitative
		  analyzing the development of the growing organs. However,
		  instance cell segmentation is still a challenging task due
		  to the quality of the image and the fine-scale structure.
		  Any small leakage in the boundary prediction can merge
		  different cells together, thereby damaging the global
		  structure of the image. In this paper, we propose an
		  end-to-end topology-aware 3D segmentation method for plant
		  tissues. The strength of the method is that it takes care
		  of the 3D topology of segmented structures. The keystone is
		  a training phase and a new topology-aware loss - the
		  CavityLoss - that are able to help the network to focus on
		  the topological errors to fix them during the learning
		  phase. The evaluation of our method on both fixed and live
		  plant organ datasets shows that our method outperforms
		  state-of-the-art methods (and contrary to state-of-the-art
		  methods, does not require any post-processing stage). The
		  code of CavityLoss is freely available at
		  \url{https://github.com/onvungocminh/CavityLoss}}
}