Fast and automatic segmentation of pulmonary lobes from chest CT using a progressive dense V-network

ABSTRACT

Automatic, reliable lobe segmentation is crucial to the diagnosis, assessment, and quantification of pulmonary diseases. Existing pulmonary lobe segmentation techniques are prohibitively slow, undesirably rely on prior (airway/vessel) segmentation, and/or require user interactions for optimal results. We introduce a reliable, fast, and fully automated lung lobe segmentation method based on a Progressive Dense V-Network (PDV-Net). The proposed method can segment lung lobes in one forward pass of the network, with an average runtime of 2 seconds using a single Nvidia Titan XP GPU. An extensive robustness analysis of our method demonstrates reliable lobe segmentation of both healthy and pathological lungs in CT images acquired by scanners from different vendors, across various CT scan protocols and acquisition parameters.

KEYWORDS:

• Lung lobe segmentation
• CT
• progressive dense V-Net
• fissure
• 3D CNN

Acknowledgments

The authors are thankful to Gerard Nguyen, MD, of the Washington University School of Medicine, for his extensive support and assistance with the annotation of lung lobes in the LIDC cases. A. Imran and A. Hatamizadeh were supported in part by an unrestricted gift to UCLA from VoxelCloud, Inc.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. At the time of the original conference paper (Imran et al. 2018) submission, no prior work had used 3D CNN-based models for lung lobe segmentation.

Additional information

Funding

This work was supported by VoxelCloud, Inc.

Notes on contributors

Abdullah-Al-Zubaer Imran

Abdullah-Al-Zubaer Imran is a PhD candidate in the Computer Science Department at the University of California, Los Angeles. He received his MS degree in Computer Science from Delaware State University in 2016. His research is focused on machine learning, generative modeling, computer vision, and medical image analysis.

Ali Hatamizadeh

Ali Hatamizadeh is a PhD candidate in the Computer Science Department at the University of California, Los Angeles. He received MS degrees in Mechanical Engineering in 2017 and Computer Science in 2019, both from UCLA. He is the recipient of the 2018 UCLA Henry Samueli School of Engineering and Applied Science Edward K. Rice Outstanding Masters Student Award.  His research interests include computer vision, computational medical imaging, and artificial intelligence in medicine.

Shilpa P. Ananth

Shilpa P. Ananth is a Machine Learning Engineer at VoxelCloud, Inc. She received her Master's degree in Biomedical Engineering from Carnegie Mellon University in 2017. Her research interests include deep learning and computer vision for medical imaging. At VoxelCloud, she is currently working on the comprehensive image analysis of Fundus Images.

Xiaowei Ding

Xiaowei Ding is Co-Founder and Chief Executive Officer at VoxelCloud, Inc. He is also a Research Assistant Professor of Computer Science at the University of California, Los Angeles. He received his PhD degree in Computer Science from UCLA in 2015. He has published papers on cardiopulmonary image analysis and deep learning applied to various medical imaging modalities.

Nima Tajbakhsh

Nima Tajbakhsh is the Lead Scientist at VoxelCloud, Inc. He received his PhD degree in Biomedical Informatics at Arizona State University in 2015. His research interests lie in the development of innovative deep learning solutions for medical image analysis from various imaging modalities, particularly with minimal annotations. His work has led to more than 40 peer-reviewed publications and 10 US patents.

Demetri Terzopoulos

Demetri Terzopoulos is a Chancellor's Professor of Computer Science at the University of California, Los Angeles, where he holds the rank of Distinguished Professor and directs the UCLA Computer Graphics & Vision Laboratory. He is also Co-Founder and Chief Science Officer of VoxelCloud, Inc. He is a fellow of the ACM, IEEE, Royal Society of Canada, and Royal Society of London. He received his PhD degree in Artificial Intelligence from MIT in 1984. He has (co-)authored more than 400 published research papers and several volumes, primarily in computer graphics, computer vision, medical imaging, computer-aided design, and artificial intelligence/life.