AssemblyNet: A Novel Deep Decision-Making Process for Whole Brain MRI Segmentation

被引:9
作者
Coupe, Pierrick [1 ]
Mansencal, Boris [1 ]
Clement, Michael [1 ]
Giraud, Remi [2 ]
de Senneville, Baudouin Denis [3 ]
Vinh-Thong Ta [1 ]
Lepetit, Vincent [1 ]
Manjon, Jose V. [4 ]
机构
[1] Univ Bordeaux, CNRS, Bordeaux INP, LABRI,UMR5800, F-33400 Talence, France
[2] Univ Bordeaux, Bordeaux INP, CNRS, IMS,UMR 5218, F-33400 Talence, France
[3] Univ Bordeaux, CNRS, IMB, UMR 5251, F-33400 Talence, France
[4] Univ Politecn Valencia, ITACA, Valencia 46022, Spain
来源
MEDICAL IMAGE COMPUTING AND COMPUTER ASSISTED INTERVENTION - MICCAI 2019, PT III | 2019年 / 11766卷
关键词
Whole brain segmentation; CNN; Ensemble learning; Transfer learning; Multiscale framework;
D O I
10.1007/978-3-030-32248-9_52
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Whole brain segmentation using deep learning (DL) is a very challenging task since the number of anatomical labels is very high compared to the number of available training images. To address this problem, previous DL methods proposed to use a global convolution neural network (CNN) or few independent CNNs. In this paper, we present a novel ensemble method based on a large number of CNNs processing different overlapping brain areas. Inspired by parliamentary decision-making systems, we propose a framework called AssemblyNet, made of two "assemblies" of U-Nets. Such a parliamentary system is capable of dealing with complex decisions and reaching a consensus quickly. AssemblyNet introduces sharing of knowledge among neighboring U-Nets, an "amendment" procedure made by the second assembly at higher-resolution to refine the decision taken by the first one, and a final decision obtained by majority voting. When using the same 45 training images, AssemblyNet outperforms global U-Net by 28% in terms of the Dice metric, patch-based joint label fusion by 15% and SLANT-27 by 10%. Finally, AssemblyNet demonstrates high capacity to deal with limited training data to achieve whole brain segmentation in practical training and testing times.
引用
收藏
页码:466 / 474
页数:9
相关论文
共 19 条
[1]  
[Anonymous], 2019, IEEE PHOTON CONF
[2]  
[Anonymous], 2016, ADV NEURAL INFORM PR
[3]   A reproducible evaluation of ANTs similarity metric performance in brain image registration [J].
Avants, Brian B. ;
Tustison, Nicholas J. ;
Song, Gang ;
Cook, Philip A. ;
Klein, Arno ;
Gee, James C. .
NEUROIMAGE, 2011, 54 (03) :2033-2044
[4]   Design and construction of a realistic digital brain phantom [J].
Collins, DL ;
Zijdenbos, AP ;
Kollokian, V ;
Sled, JG ;
Kabani, NJ ;
Holmes, CJ ;
Evans, AC .
IEEE TRANSACTIONS ON MEDICAL IMAGING, 1998, 17 (03) :463-468
[5]  
de Brebisson Alexandre, 2015, 2015 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), P20, DOI 10.1109/CVPRW.2015.7301312
[6]   3D whole brain segmentation using spatially localized atlas network tiles [J].
Huo, Yuankai ;
Xu, Zhoubing ;
Xiong, Yunxi ;
Aboud, Katherine ;
Parvathaneni, Prasanna ;
Bao, Shunxing ;
Bermudez, Camilo ;
Resnick, Susan M. ;
Cutting, Laurie E. ;
Landman, Bennett A. .
NEUROIMAGE, 2019, 194 :105-119
[7]  
Izmailov Pavel, 2018, ARXIV180305407
[8]   CANDIShare: A Resource for Pediatric Neuroimaging Data [J].
Kennedy, David N. ;
Haselgrove, Christian ;
Hodge, Steven M. ;
Rane, Pallavi S. ;
Makris, Nikos ;
Frazier, Jean A. .
NEUROINFORMATICS, 2012, 10 (03) :319-322
[9]  
Manjon J.V., 2014, IJBI, V2014, P10
[10]   Robust MRI brain tissue parameter estimation by multistage outlier rejection [J].
Manjon, Jose V. ;
Tohka, Jussi ;
Garcia-Marti, Gracian ;
Carbonell-Caballero, Jose ;
Lull, Juan J. ;
Marti-Bonmati, Luis ;
Robles, Montserrat .
MAGNETIC RESONANCE IN MEDICINE, 2008, 59 (04) :866-873