Abstract
The purpose of this study was to compare a manual bronchoscopic navigation technique, the direct oblique method (DOM), with conventional virtual bronchoscopic navigation software in terms of bronchial identification ability involving reconstruction of a whole bronchial tree from identical CT images. A whole bronchial tree was drawn using manual bronchial recognition with the DOM. The tree was compared with that reconstructed by SYNAPSE VINCENT bronchoscopic navigation-dedicated software. The number of bronchial generations at each terminal tip was then compared between the two approaches. Physicians spent 20 h tracing all bronchi on CT scan images and obtained a bronchial tree. The hand-made bronchial tree had five times the number of tips as that reconstructed by automatic bronchial recognition (1482 vs. 279 tips, respectively). The number of bronchial generations prior to each terminal tip was larger with the DOM than with VINCENT (median, 10; interquartile range (IQR), 9–11 vs. median, 5; IQR, 5–7, respectively; p-value < 0.001). Using the CT image data in this case, manual bronchial recognition with the DOM identified more bronchi than automatic bronchial recognition. This result implies that manual bronchial recognition is a valid basis for detailed bronchoscopic navigation analysis.
Similar content being viewed by others
Introduction
Bronchial segmentation on CT images is fundamental to respiratory medicine, including navigated bronchoscopy and the diagnosis of airway diseases. Although software-based automatic bronchial identification is improving, it still cannot identify all bronchi on CT images1. Recently, we developed the direct oblique method (DOM), a manual but ultra-precise bronchoscopic navigation technique. However, it remains unclear how the DOM and conventional virtual bronchoscopic navigation compare regarding their bronchial identification ability.
Methods
A full-chest, thin-slice CT scan (resolution, 512 × 512; slice thickness, 0.625 mm; and slice distance, 0.625 mm) of a patient who underwent chest screening in Osaka University Hospital was analyzed. The whole bronchial tree was manually drawn by direct bronchial recognition on CT images and visualized on a Ziostation2 CT bronchoscopy navigation system (referred to as “Ziostation2”; Ziosoft, Tokyo, Japan). All bronchi and bifurcations were handwritten one by one throughout the lungs. Transverse and longitudinal cross-sectional oblique CT images were used to identify each of the bronchi, just as with the DOM. This procedure is shown in the Supplementary Video. The resultant bronchial tree was compared with that derived using automatic bronchial recognition by SYNAPSE VINCENT bronchoscopic navigation-dedicated software (referred to as “VINCENT-BFsim”; Fujifilm Medical, Tokyo, Japan).
To quantify bronchial identification ability, the number of bronchial generations at each terminal tip was counted. We referred to segmental bronchi as the second generation, subsegmental bronchi as the third generation, and so forth.
The number of bronchial generations prior to each terminal tip was compared with Student’s t-test using R ver. 4.0.2 software (available at http://www.R-project.org; R Foundation for Statistical Computing, Vienna, Austria). A p-value of < 0.05 was considered statistically significant.
This study was approved as Protocol #864 by the Research Ethics Committee of Osaka University. The research was performed in accordance with the Declaration of Helsinki. Written informed consent was obtained from the patient whose CT scan images were used in this study.
Results
Physicians spent 20 h tracing all bronchi in CT scan images to obtain a bronchial tree. On the other hand, VINCENT-BFsim extracted the bronchial tree in less than 2 min. The bronchial tree identified by the DOM on Ziostation2 had five times the number of tips as that depicted by VINCENT-BFsim (1482 vs. 279 tips, respectively) (Fig. 1). The number of bronchial generations prior to each terminal tip was significantly greater with the DOM than with VINCENT-BSsim (median, 10; interquartile range: IQR, 9–11 vs. median, 5; IQR, 5–7, respectively; p-value < 0.001) (Fig. 2), while the maximum number was 14 for the DOM (Fig. 3) and 10 for VINCENT-BFsim. These results demonstrated that the DOM identified more bronchi than the automatic system.
Discussion
This study demonstrated that manual handwriting of bronchi on CT images remains a valuable approach today, since many bronchi were not identified by a current software-based automatic bronchial identification method. To quantify the performance of bronchial identification, we compared the number of tips and bronchial generations prior to each terminal tip of the bronchial tree. Although only one bronchial tree was drawn due to the strenuous work involved, the DOM restored the bronchial tree more accurately than the automatic method in this case.
There have been significant advances in CT-based bronchial segmentation methods and bronchoscopic navigation. Conventional manual navigation techniques sometimes yield incorrect and insufficient information because they are based on bronchial recognition on axial CT-images with mental reconstruction of three-dimensional structures2. Bronchoscopic navigation systems with automatic bronchial segmentation improve diagnostic yield3,4,5,6,7, but their ability to recognize bronchi is still inadequate and varies between software systems. For example, when compared to LungPoint (Broncus Medical, Mountain View, CA), a globally used bronchoscopic navigation system, VINCENT-BFsim has better bronchial segmentation performance1,8 but it still cannot identify all bronchi on CT images near the target lesion1. The DOM is a manual navigation technique, which makes it possible to use all bronchi on CT images for bronchoscopic navigation1. However, unlike in conventional manual analysis, the DOM does not require mental reconstruction because oblique sections are used instead of axial images. In addition, bronchi are carefully evaluated in both the transverse and longitudinal cross-sections of each bronchus1. Since a limited area of analysis is needed for bronchoscopy preparation, the DOM takes only about 5 min per case1, which is almost equivalent to VINCENT-BFsim.
We propose the efficacy of bronchoscopic navigation systems such as Ziostation2 in which automatic bronchial identification and precise handwriting are seamlessly fused. Further study is needed to determine whether such precise navigation improves diagnostic rates.
References
Miyake, K. et al. The direct oblique method: A new gold standard for bronchoscopic navigation that is superior to automatic methods. J. Bronchol. Interv. Pulmonol. 25, 305–314. https://doi.org/10.1097/LBR.0000000000000512 (2018).
Gibbs, J. D., Graham, M. W. & Higgins, W. E. 3D MDCT-based system for planning peripheral bronchoscopic procedures. Comput. Biol. Med. 39, 266–279. https://doi.org/10.1016/j.compbiomed.2008.12.012 (2009).
Tamiya, M. et al. Diagnostic yield of combined bronchoscopy and endobronchial ultrasonography, under LungPoint guidance for small peripheral pulmonary lesions. Respirology 18, 834–839. https://doi.org/10.1111/resp.12095 (2013).
Ishida, T. et al. Virtual bronchoscopic navigation combined with endobronchial ultrasound to diagnose small peripheral pulmonary lesions: A randomised trial. Thorax 66, 1072–1077. https://doi.org/10.1136/thx.2010.145490 (2011).
Eberhardt, R. et al. LungPoint: A new approach to peripheral lesions. J. Thorac. Oncol. 5, 1559–1563. https://doi.org/10.1097/jto.0b013e3181e8b308 (2010).
Dolina, M. Y. et al. Interbronchoscopist variability in endobronchial path selection. Chest 133, 897–905. https://doi.org/10.1378/chest.07-2540 (2008).
Asano, F. et al. Virtual bronchoscopic navigation combined with ultrathin bronchoscopy. A randomized clinical trial. Am. J. Respir. Crit. Care Med. 188, 327–333. https://doi.org/10.1164/rccm.201211-2104OC (2013).
Kitamura, A. et al. Reproducibility of peripheral branches in virtual bronchoscopic navigation using VINCENT and LungPoint software for peripheral lung lesions. Respir. Investig. 59, 772–776. https://doi.org/10.1016/j.resinv.2021.04.006 (2021).
Acknowledgements
This study was supported in part by Ziosoft Inc. (Tokyo, Japan), which conducted the CT image analysis.
Funding
This work was supported by the Japan Society for the Promotion of Science KAKENHI [JP18H05282 to A.K.]; Center of Innovation program from the Ministry of Education, Culture, Sports, Science and Technology of Japan [to A.K.]; Japan Agency for Medical Research and Development [J200705023, J200705710, J200705049, JP18cm016335, and JP18cm059042 to A.K.]; Kansai Economic Federation grant [to A.K.]; and Mitsubishi Zaidan grants [to A.K.].
Author information
Authors and Affiliations
Contributions
S.T. and K.M. performed the study design, analysis, and writing of the manuscript. A.K. provided the final approval of the version to be published.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher's note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary Information
Supplementary Video 1.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Takata, S., Miyake, K. & Kumanogoh, A. The superiority of manual over automated methods in identifying bronchial trees on identical CT images. Sci Rep 12, 5416 (2022). https://doi.org/10.1038/s41598-022-09401-8
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-022-09401-8
- Springer Nature Limited