A Squeeze Convolutional Network For MRI Right Ventricle Segmentation

Segmentation of the right ventricle (RV) is difficult due to the variable shape and fuzziness of RV boundaries. In this paper, we propose an effective squeeze Convolutional Network for RV segmentation. A squeeze expand attention (SEA) module is constructed by combining squeezing, expanding, and adaptive weighting of features to generate feature maps for prediction. Our SEA module is introduced into a U-shaped network architecture to extract RV features in the coding layers and make end-to-end decisions in the decoding layers. Besides, a loss function constrained by RV shape constraints is proposed in our network to attenuate abnormal prediction results. Our proposed network is with fewer parameters and is rapid in training. Experimental results on public datasets show that our proposed network achieves more accurate segmentation accuracy compared with the existing methods significantly in terms of accuracy and speed.

[1]  Georgios B. Giannakis,et al.  Fully Automatic Segmentation of the Right Ventricle Via Multi-Task Deep Neural Networks , 2018, 2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP).

[2]  Edward J. Delp,et al.  Locating Objects Without Bounding Boxes , 2018, 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).

[3]  Daniel Rueckert,et al.  Combining Deep Learning and Shape Priors for Bi-Ventricular Segmentation of Volumetric Cardiac Magnetic Resonance Images , 2018, ShapeMI@MICCAI.

[4]  Nicholas Ayache,et al.  3-D Consistent and Robust Segmentation of Cardiac Images by Deep Learning With Spatial Propagation , 2018, IEEE Transactions on Medical Imaging.

[5]  François Chollet,et al.  Xception: Deep Learning with Depthwise Separable Convolutions , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

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

[7]  Bertil Schmidt,et al.  $ν$-net: Deep Learning for Generalized Biventricular Cardiac Mass and Function Parameters , 2017, ArXiv.

[8]  Olga V. Senyukova,et al.  Right Ventricle Segmentation in Cardiac MR Images Using U-Net with Partly Dilated Convolution , 2018, ICANN.

[9]  Sébastien Ourselin,et al.  Multi-atlas Propagation Whole Heart Segmentation from MRI and CTA Using a Local Normalised Correlation Coefficient Criterion , 2013, FIMH.

[10]  Jiaming Guo,et al.  Deep patch-wise colorization model for grayscale images , 2016, SIGGRAPH Asia Technical Briefs.

[11]  Hamid Jafarkhani,et al.  Automatic segmentation of the right ventricle from cardiac MRI using a learning‐based approach , 2017, Magnetic resonance in medicine.

[12]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[13]  Ben Glocker,et al.  Semi-supervised Learning for Network-Based Cardiac MR Image Segmentation , 2017, MICCAI.

[14]  Dong-Hyun Kim,et al.  Fully coupled aero-hydrodynamic analysis of a semi-submersible FOWT using a dynamic fluid body interaction approach , 2016 .

[15]  Jitendra Malik,et al.  Region-Based Convolutional Networks for Accurate Object Detection and Segmentation , 2016, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[16]  Roberto Cipolla,et al.  Deep Roots: Improving CNN Efficiency with Hierarchical Filter Groups , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[17]  Damini Dey,et al.  Cardiac imaging: working towards fully-automated machine analysis & interpretation , 2017, Expert review of medical devices.

[18]  Enhua Wu,et al.  Squeeze-and-Excitation Networks , 2017, IEEE Transactions on Pattern Analysis and Machine Intelligence.