A global benchmark of algorithms for segmenting the left atrium from late gadolinium-enhanced cardiac magnetic resonance imaging

Segmentation of medical images, particularly late gadolinium-enhanced magnetic resonance imaging (LGE-MRI) used for visualizing diseased atrial structures, is a crucial first step for ablation treatment of atrial fibrillation. However, direct segmentation of LGE-MRIs is challenging due to the varying intensities caused by contrast agents. Since most clinical studies have relied on manual, labor-intensive approaches, automatic methods are of high interest, particularly optimized machine learning approaches. To address this, we organized the 2018 Left Atrium Segmentation Challenge using 154 3D LGE-MRIs, currently the world's largest atrial LGE-MRI dataset, and associated labels of the left atrium segmented by three medical experts, ultimately attracting the participation of 27 international teams. In this paper, extensive analysis of the submitted algorithms using technical and biological metrics was performed by undergoing subgroup analysis and conducting hyper-parameter analysis, offering an overall picture of the major design choices of convolutional neural networks (CNNs) and practical considerations for achieving state-of-the-art left atrium segmentation. Results show that the top method achieved a Dice score of 93.2% and a mean surface to surface distance of 0.7 mm, significantly outperforming prior state-of-the-art. Particularly, our analysis demonstrated that double sequentially used CNNs, in which a first CNN is used for automatic region-of-interest localization and a subsequent CNN is used for refined regional segmentation, achieved superior results than traditional methods and machine learning approaches containing single CNNs. This large-scale benchmarking study makes a significant step towards much-improved segmentation methods for atrial LGE-MRIs, and will serve as an important benchmark for evaluating and comparing the future works in the field. Furthermore, the findings from this study can potentially be extended to other imaging datasets and modalities, having an impact on the wider medical imaging community.

[1]  Aimin Hao,et al.  Automatic 3D Atrial Segmentation from GE-MRIs Using Volumetric Fully Convolutional Networks , 2018, STACOM@MICCAI.

[2]  Cristiana Corsi,et al.  A Semantic-Wise Convolutional Neural Network Approach for 3-D Left Atrium Segmentation from Late Gadolinium Enhanced Magnetic Resonance Imaging , 2018, STACOM@MICCAI.

[3]  Nassir Marrouche,et al.  Dark regions of no-reflow on late gadolinium enhancement magnetic resonance imaging result in scar formation after atrial fibrillation ablation. , 2011, Journal of the American College of Cardiology.

[4]  Sailesh Conjeti,et al.  Machine Learning for Medical Image Reconstruction , 2018, Lecture Notes in Computer Science.

[5]  Sergey Ioffe,et al.  Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning , 2016, AAAI.

[6]  George Papandreou,et al.  Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation , 2018, ECCV.

[7]  Jichao Zhao,et al.  Fully Automatic Left Atrium Segmentation From Late Gadolinium Enhanced Magnetic Resonance Imaging Using a Dual Fully Convolutional Neural Network , 2019, IEEE Transactions on Medical Imaging.

[8]  Heye Zhang,et al.  Simultaneous left atrium anatomy and scar segmentations via deep learning in multiview information with attention , 2020, Future Gener. Comput. Syst..

[9]  Nathalie Harder,et al.  An Objective Comparison of Cell Tracking Algorithms , 2017, Nature Methods.

[10]  Daniel Rueckert,et al.  Multi-Task Learning for Left Atrial Segmentation on GE-MRI , 2018, STACOM@MICCAI.

[11]  Seunghoon Hong,et al.  Learning Deconvolution Network for Semantic Segmentation , 2015, 2015 IEEE International Conference on Computer Vision (ICCV).

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

[13]  Xin Yang,et al.  Deep Learning Techniques for Automatic MRI Cardiac Multi-Structures Segmentation and Diagnosis: Is the Problem Solved? , 2018, IEEE Transactions on Medical Imaging.

[14]  Jürgen Weese,et al.  Benchmark for Algorithms Segmenting the Left Atrium From 3D CT and MRI Datasets , 2015, IEEE Transactions on Medical Imaging.

[15]  Richard James Housden,et al.  Algorithms for left atrial wall segmentation and thickness – Evaluation on an open-source CT and MRI image database , 2018, Medical Image Anal..

[16]  Sandy Engelhardt,et al.  Segmentation of the Left Atrium from 3D Gadolinium-Enhanced MR Images with Convolutional Neural Networks , 2018, STACOM@MICCAI.

[17]  Li Fei-Fei,et al.  ImageNet: A large-scale hierarchical image database , 2009, CVPR.

[18]  Trevor Darrell,et al.  Fully Convolutional Networks for Semantic Segmentation , 2017, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[19]  Yi Gao,et al.  Automatic Segmentation of the Left Atrium From MR Images via Variational Region Growing With a Moments-Based Shape Prior , 2013, IEEE Transactions on Image Processing.

[20]  Hervé Delingette,et al.  Automatically Segmenting the Left Atrium from Cardiac Images Using Successive 3D U-Nets and a Contour Loss , 2018, STACOM@MICCAI.

[21]  Guang Yang,et al.  Atrial scar quantification via multi-scale CNN in the graph-cuts framework , 2019, Medical Image Anal..

[22]  Xin Yang,et al.  Pyramid Network with Online Hard Example Mining for Accurate Left Atrium Segmentation , 2018, STACOM@MICCAI.

[23]  R. J. van der Geest,et al.  Fully automatic segmentation of left atrium and pulmonary veins in late gadolinium‐enhanced MRI: Towards objective atrial scar assessment , 2016, Journal of magnetic resonance imaging : JMRI.

[24]  Xin Yang,et al.  Combating Uncertainty with Novel Losses for Automatic Left Atrium Segmentation , 2018, STACOM@MICCAI.

[25]  Andreas K. Maier,et al.  Dilated Convolutions in Neural Networks for Left Atrial Segmentation in 3D Gadolinium Enhanced-MRI , 2018, STACOM@MICCAI.

[26]  Francisco Alarcón,et al.  Postprocedural LGE‐CMR comparison of laser and radiofrequency ablation lesions after pulmonary vein isolation , 2018, Journal of cardiovascular electrophysiology.

[27]  Lei Li,et al.  Left Atrial Segmentation Combining Multi-atlas Whole Heart Labeling and Shape-Based Atlas Selection , 2018, STACOM@MICCAI.

[28]  Jichao Zhao,et al.  Statistical Atlases and Computational Models of the Heart. Atrial Segmentation and LV Quantification Challenges , 2018, Lecture Notes in Computer Science.

[29]  Christopher McGann,et al.  Atrial Fibrillation Ablation Outcome Is Predicted by Left Atrial Remodeling on MRI , 2014, Circulation. Arrhythmia and electrophysiology.

[30]  Pablo Lamata,et al.  V-FCNN: Volumetric Fully Convolution Neural Network For Automatic Atrial Segmentation , 2018, STACOM@MICCAI.

[31]  Cong Yan,et al.  Deep Learning Based Method for Left Atrial Segmentation in GE-MRI , 2018, STACOM@MICCAI.

[32]  Joshua J. E. Blauer,et al.  Detection and Quantification of Left Atrial Structural Remodeling With Delayed-Enhancement Magnetic Resonance Imaging in Patients With Atrial Fibrillation , 2009, Circulation.

[33]  Hubert Cochet,et al.  High-resolution three-dimensional late gadolinium-enhanced cardiac magnetic resonance imaging to identify the underlying substrate of ventricular arrhythmia , 2017, Europace : European pacing, arrhythmias, and cardiac electrophysiology : journal of the working groups on cardiac pacing, arrhythmias, and cardiac cellular electrophysiology of the European Society of Cardiology.

[34]  Qiong Wang,et al.  Attention Based Hierarchical Aggregation Network for 3D Left Atrial Segmentation , 2018, STACOM@MICCAI.

[35]  Paul M. Thompson,et al.  Multimodal Brain Image Analysis and Mathematical Foundations of Computational Anatomy: 4th International Workshop, MBIA 2019, and 7th International Workshop, MFCA 2019, Held in Conjunction with MICCAI 2019, Shenzhen, China, October 17, 2019, Proceedings , 2019, MBIA/MFCA@MICCAI.

[36]  Jichao Zhao,et al.  Three‐dimensional Integrated Functional, Structural, and Computational Mapping to Define the Structural “Fingerprints” of Heart‐Specific Atrial Fibrillation Drivers in Human Heart Ex Vivo , 2017, Journal of the American Heart Association.

[37]  Ross T. Whitaker,et al.  ShapeCut: Bayesian surface estimation using shape‐driven graph , 2017, Medical Image Anal..

[38]  Avan Suinesiaputra,et al.  Challenges of Cardiac Image Analysis in Large-Scale Population-Based Studies , 2015, Current Cardiology Reports.

[39]  Qian Tao,et al.  Fully Automated Left Atrium Cavity Segmentation from 3D GE-MRI by Multi-atlas Selection and Registration , 2018, STACOM@MICCAI.

[40]  Zhou Zhao,et al.  Left Atrial Segmentation in a Few Seconds Using Fully Convolutional Network and Transfer Learning , 2018, STACOM@MICCAI.

[41]  Nazem Akoum,et al.  The Spatial Distribution of Late Gadolinium Enhancement of Left Atrial Magnetic Resonance Imaging in Patients With Atrial Fibrillation. , 2018, JACC. Clinical electrophysiology.

[42]  Heng Liu,et al.  Segmentation and visualization of left atrium through a unified deep learning framework , 2020, International Journal of Computer Assisted Radiology and Surgery.

[43]  Kawal S. Rhode,et al.  Convolutional Neural Networks for Segmentation of the Left Atrium from Gadolinium-Enhancement MRI Images , 2018, STACOM@MICCAI.

[44]  Jichao Zhao,et al.  Novel application of 3D contrast-enhanced CMR to define fibrotic structure of the human sinoatrial node in vivo , 2017, European heart journal cardiovascular Imaging.