Book chapter
A Novel Iterative Method for Airway Tree Segmentation from CT Imaging Using Multiscale Leakage Detection
Computer Vision – ACCV 2016 Workshops, pp.46-60
Lecture Notes in Computer Science, Springer International Publishing
03/16/2017
DOI: 10.1007/978-3-319-54526-4_4
Abstract
Computed tomography (CT)-based metrics of airway phenotypes, wall-thickness, and other morphological features are increasingly being used in large multi-center lung studies involving many hundreds or thousands of image datasets. There is an unmet need for a fully reliable, automated algorithm for CT-based segmentation of airways. State-of-the-art methods require a post-editing step, which is time consuming when several thousands of image data sets need to be reviewed and edited. In this paper, we present a novel iterative algorithm for CT-based segmentation of airway trees. Early testing suggests that the method requires no editing to extract a set of airway segments along a standardized set of bronchial paths extending two generations beyond the segmental airways. It uses simple intensity-based connectivity and new leakage detection and volume freezing algorithms to iteratively grow an airway tree. It starts with an initial, automatically determined seed inside the trachea and a conservative threshold; applies region growing and generates a leakage-corrected segmentation; freezes the segmented volume; and shifts the threshold toward a more generous value for the next iteration until a convergence occurs. The method was applied on chest CT scans of fifteen normal non-smoking subjects. Airway segmentation results were compared with manually edited results, and branch level accuracy of the new segmentation method was examined along five standardized segmental airway paths and continuing to two generations beyond the segmental paths. The method successfully detected all branches up to two generations beyond the five segmental airway paths with no visual leakages.
Details
- Title: Subtitle
- A Novel Iterative Method for Airway Tree Segmentation from CT Imaging Using Multiscale Leakage Detection
- Creators
- Syed Ahmed Nadeem - University of IowaDakai Jin - University of IowaEric A Hoffman - University of IowaPunam K Saha - University of Iowa
- Resource Type
- Book chapter
- Publication Details
- Computer Vision – ACCV 2016 Workshops, pp.46-60
- Publisher
- Springer International Publishing; Cham
- Series
- Lecture Notes in Computer Science
- DOI
- 10.1007/978-3-319-54526-4_4
- eISSN
- 1611-3349
- ISSN
- 0302-9743
- Language
- English
- Date published
- 03/16/2017
- Academic Unit
- Electrical and Computer Engineering; Radiology; Roy J. Carver Department of Biomedical Engineering; Internal Medicine
- Record Identifier
- 9984197526302771
Metrics
10 Record Views