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

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.

[1]  Tanveer F. Syeda-Mahmood,et al.  3D Segmentation with Exponential Logarithmic Loss for Highly Unbalanced Object Sizes , 2018, MICCAI.

[2]  Thomas Brox,et al.  U-Net: Convolutional Networks for Biomedical Image Segmentation , 2015, MICCAI.

[3]  Hongyi Zhang,et al.  mixup: Beyond Empirical Risk Minimization , 2017, ICLR.

[4]  D. Louis Collins,et al.  Nonlocal Intracranial Cavity Extraction , 2014, Int. J. Biomed. Imaging.

[5]  Giovanni Montana,et al.  Deep neural networks for anatomical brain segmentation , 2015, 2015 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW).

[6]  John G. Csernansky,et al.  Open Access Series of Imaging Studies (OASIS): Cross-sectional MRI Data in Young, Middle Aged, Nondemented, and Demented Older Adults , 2007, Journal of Cognitive Neuroscience.

[7]  Christian Wachinger,et al.  DeepNAT: Deep convolutional neural network for segmenting neuroanatomy , 2017, NeuroImage.

[8]  Mert R. Sabuncu,et al.  VoxelMorph: A Learning Framework for Deformable Medical Image Registration , 2018, IEEE Transactions on Medical Imaging.

[9]  Frazier Jean,et al.  CANDIShare: A Resource for Pediatric Neuroimaging Data , 2010 .

[10]  Zoubin Ghahramani,et al.  A Theoretically Grounded Application of Dropout in Recurrent Neural Networks , 2015, NIPS.

[11]  Brian B. Avants,et al.  N4ITK: Improved N3 Bias Correction , 2010, IEEE Transactions on Medical Imaging.

[12]  D. Louis Collins,et al.  Design and construction of a realistic digital brain phantom , 1998, IEEE Transactions on Medical Imaging.

[13]  Andrew Gordon Wilson,et al.  Averaging Weights Leads to Wider Optima and Better Generalization , 2018, UAI.

[14]  Pierrick Coupé,et al.  Author manuscript, published in "Journal of Magnetic Resonance Imaging 2010;31(1):192-203" DOI: 10.1002/jmri.22003 Adaptive Non-Local Means Denoising of MR Images with Spatially Varying Noise Levels , 2010 .

[15]  Jussi Tohka,et al.  Robust MRI brain tissue parameter estimation by multistage outlier rejection , 2008, Magnetic resonance in medicine.

[16]  Shunxing Bao,et al.  3D whole brain segmentation using spatially localized atlas network tiles , 2019, NeuroImage.

[17]  Paul A. Yushkevich,et al.  Multi-atlas segmentation with joint label fusion and corrective learning—an open source implementation , 2013, Front. Neuroinform..

[18]  Arno Klein,et al.  A reproducible evaluation of ANTs similarity metric performance in brain image registration , 2011, NeuroImage.

[19]  Nassir Navab,et al.  Error Corrective Boosting for Learning Fully Convolutional Networks with Limited Data , 2017, MICCAI.