View source for Publications/zhao.20.icpr.2

{{Publication
| published = true
| date = 2020-11-02
| authors = Zhou Zhao, Nicolas Boutry, Élodie Puybareau, Thierry Géraud
| title = Do not Treat Boundaries and Regions Differently: An Example on Heart Left Atrial Segmentation
| booktitle = Proceedings of the 25th International Conference on Pattern Recognition (ICPR)
| pages = 7447 to 7453
| address = Milan, Italy
| publisher = IEEE
| abstract = Atrial fibrillation is the most common heart rhythm disease. Due to a lack of understanding in matter of underlying atrial structures, current treatments are still not satisfying. Recently, with the popularity of deep learning, many segmentation methods based on fully convolutional networks have been proposed to analyze atrial structures, especially from late gadolinium-enhanced magnetic resonance imaging. However, two problems still occur: 1) segmentation results include the atrial- like background; 2) boundaries are very hard to segment. Most segmentation approaches design a specific network that mainly focuses on the regions, to the detriment of the boundaries. Therefore, this paper proposes an attention full convolutional network framework based on the ResNet-101 architecture, which focuses on boundaries as much as on regions. The additional attention module is added to have the network pay more attention on regions and then to reduce the impact of the misleading similarity of neighboring tissues. We also use a hybrid loss composed of a region loss and a boundary loss to treat boundaries and regions at the same time. We demonstrate the efficiency of the proposed approach on the MICCAI 2018 Atrial Segmentation Challenge public dataset.
| lrdepaper = http://www.lrde.epita.fr/dload/papers/zhao.20.icpr.2.pdf
| lrdeprojects = Olena
| lrdekeywords = Image
| lrdenewsdate = 2020-11-02
| type = inproceedings
| id = zhao.20.icpr.2
| identifier = doi:10.1109/ICPR48806.2021.9412755
| bibtex = 
@InProceedings<nowiki>{</nowiki>	  zhao.20.icpr.2,
  author	= <nowiki>{</nowiki>Zhou Zhao and Nicolas Boutry and \'Elodie Puybareau and
		  Thierry G\'eraud<nowiki>}</nowiki>,
  title		= <nowiki>{</nowiki>Do not Treat Boundaries and Regions Differently: <nowiki>{</nowiki>A<nowiki>}</nowiki>n
		  Example on Heart Left Atrial Segmentation<nowiki>}</nowiki>,
  booktitle	= <nowiki>{</nowiki>Proceedings of the 25th International Conference on
		  Pattern Recognition (ICPR)<nowiki>}</nowiki>,
  year		= 2021,
  pages		= <nowiki>{</nowiki>7447--7453<nowiki>}</nowiki>,
  month		= <nowiki>{</nowiki>January<nowiki>}</nowiki>,
  address	= <nowiki>{</nowiki>Milan, Italy<nowiki>}</nowiki>,
  publisher	= <nowiki>{</nowiki>IEEE<nowiki>}</nowiki>,
  abstract	= <nowiki>{</nowiki>Atrial fibrillation is the most common heart rhythm
		  disease. Due to a lack of understanding in matter of
		  underlying atrial structures, current treatments are still
		  not satisfying. Recently, with the popularity of deep
		  learning, many segmentation methods based on fully
		  convolutional networks have been proposed to analyze atrial
		  structures, especially from late gadolinium-enhanced
		  magnetic resonance imaging. However, two problems still
		  occur: 1) segmentation results include the atrial- like
		  background; 2) boundaries are very hard to segment. Most
		  segmentation approaches design a specific network that
		  mainly focuses on the regions, to the detriment of the
		  boundaries. Therefore, this paper proposes an attention
		  full convolutional network framework based on the
		  ResNet-101 architecture, which focuses on boundaries as
		  much as on regions. The additional attention module is
		  added to have the network pay more attention on regions and
		  then to reduce the impact of the misleading similarity of
		  neighboring tissues. We also use a hybrid loss composed of
		  a region loss and a boundary loss to treat boundaries and
		  regions at the same time. We demonstrate the efficiency of
		  the proposed approach on the MICCAI 2018 Atrial
		  Segmentation Challenge public dataset.<nowiki>}</nowiki>,
  doi		= <nowiki>{</nowiki>10.1109/ICPR48806.2021.9412755<nowiki>}</nowiki>
<nowiki>}</nowiki>

}}