iBet uBet web content aggregator. Adding the entire web to your favor.
iBet uBet web content aggregator. Adding the entire web to your favor.



Link to original content: https://pubmed.ncbi.nlm.nih.gov/21997248
Automated 3-D segmentation of lungs with lung cancer in CT data using a novel robust active shape model approach - PubMed Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2012 Feb;31(2):449-60.
doi: 10.1109/TMI.2011.2171357. Epub 2011 Oct 13.

Automated 3-D segmentation of lungs with lung cancer in CT data using a novel robust active shape model approach

Shanhui Sun  1 , Christian Bauer, Reinhard Beichel
Affiliations

Automated 3-D segmentation of lungs with lung cancer in CT data using a novel robust active shape model approach

Shanhui Sun et al. IEEE Trans Med Imaging. 2012 Feb.

Abstract

Segmentation of lungs with (large) lung cancer regions is a nontrivial problem. We present a new fully automated approach for segmentation of lungs with such high-density pathologies. Our method consists of two main processing steps. First, a novel robust active shape model (RASM) matching method is utilized to roughly segment the outline of the lungs. The initial position of the RASM is found by means of a rib cage detection method. Second, an optimal surface finding approach is utilized to further adapt the initial segmentation result to the lung. Left and right lungs are segmented individually. An evaluation on 30 data sets with 40 abnormal (lung cancer) and 20 normal left/right lungs resulted in an average Dice coefficient of 0.975±0.006 and a mean absolute surface distance error of 0.84±0.23 mm, respectively. Experiments on the same 30 data sets showed that our methods delivered statistically significant better segmentation results, compared to two commercially available lung segmentation approaches. In addition, our RASM approach is generally applicable and suitable for large shape models.

PubMed Disclaimer

Figures

Fig. 1
Fig. 1
Segmentation of a lung with cancer using a conventional approach. (a) Axial CT image showing normal right and cancerous left lung tissue. (b) Corresponding segmentation result generated with a conventional lung segmentation method. Segmentation errors are indicated by arrows.
Fig. 2
Fig. 2
Overview of our model-based segmentation approach.
Fig. 3
Fig. 3
Outline of main rib detection processing steps. (a) Volume rendering of the input thorax CT data truncated to a gray-value range between 0 and 500 HU. (b) Volume rendering of Frangi’s “vesselness measure” [18] computed at a scale of σ = 5 mm and (c) corresponding centerlines of rib candidates (Section II-B1). Note that many responses from vessels (e.g., aorta) can be found in (b) and (c), because the CT image is contrast enhanced. Centerlines from vessels and other non-rib structures are removed in subsequent rib detection steps. Output of first (d) and second (e) rib clustering/detection stage (Section II-B2).
Fig. 4
Fig. 4
Visualization of eigenvector patterns utilized for rib detection. Eigenvectors are shown for ribs [(a), (b)] and the aortic arch (c).
Fig. 5
Fig. 5
Robust shape pattern learning. A random sampling process is repeated l-times (rows). In each random sampling process, k shape subsets are derived from all n training shapes and utilized to generated point subset distribution models.
Fig. 6
Fig. 6
Histogram of component values of verr and automatically derived threshold utilized to detect outlier components.
Fig. 7
Fig. 7
Comparison of the Dice coefficient of standard ASM, robust ASM (RASM), and proposed (RASM+OSF) lung segmentation approaches for (a) left and (b) right lungs. Note that boxplots for normal (N) and diseased (D) lungs are shown separately.
Fig. 8
Fig. 8
Segmentation result for the example shown in Fig. 1(a). (a) Reference segmentation and (b) proposed segmentation approach.
Fig. 9
Fig. 9
Boxplots of the Dice coefficient for P1, P2, and our approach.
Fig. 10
Fig. 10
Examples of segmentation results on three different data sets (rows). (a) , (f), (k) Results for method P1. (b)–(d), (g)–(i), (l)–(n) Results generated with method P2 based on three different template initializations. (e), (j), (o) Results of the proposed automated approach (RASM+OSF).
Fig. 11
Fig. 11
Comparison between conventional and proposed lung segmentation methods. (a) The conventional method leaks into the gas filled colon. (b) Our method provides a correct segmentation.
Fig. 12
Fig. 12
Performance comparison between (a) standard ASM and (b) RASM matching (without optimal surface finding step). The RASM delivers a better match for normal and diseased lungs.
Fig. 13
Fig. 13
Segmentation of an incomplete lung CT data set; the top portion was not scanned. (a) Standard ASM. (b) Robust ASM (without optimal surface finding step). Note that the standard and robust ASM are not aware of the spatial extent of the data set, because of the clamping of gradient values to the boundary (Section II-C1). Surfaces outside of the data set were clipped after the segmentation process was completed.
Fig. 14
Fig. 14
Examples of RASM segmentation results (without optimal surface finding step). (a) Case with small right lung and (b) pleural effusion in left lung. See text in Section V for details.
Fig. 15
Fig. 15
Example segmentation results of lung with idiopathic pulmonary fibrosis. (a) A manual reference segmentation. (b) Result of a conventional segmentation method. (c) Preliminary segmentation result of our approach (RASM+OSF).
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Armato SG, Sensakovic WF. Automated lung segmentation for thoracic CT. Acad. Radiol. 2004;vol. 11(no. 9):1011–1021. - PubMed
    1. Leader JK, Zheng B, Rogers RM, Sciurba FC, Perez A, Chapman BE, Patel S, Fuhrman CR, Gur D. Automated lung segmentation in X-ray computed tomography. Acad. Radiol. 2003;vol. 10(no. 11):1224–1236. - PubMed
    1. Silva A, Silva JS, Santos BS, Ferreira C. Fast pulmonary contour extraction in X-ray CT images: A methodology and quality assessment. Proc. SPIE Med. Imag. 2001;vol. 4321:216–224.
    1. Hu S, Hoffman EA, Reinhardt JM. Automatic lung segmentation for accurate quantitation of volumetric X-ray CT image. IEEE Trans. Med. Imag. 2001 Jun.vol. 20(no. 6):490–498. - PubMed
    1. Hoffman EA, Ritman EL. Effect of body orientation on regional lung expansion in dog and sloth. J. Appl. Physiol. 1985;vol. 59(no. 2):481–491. - PubMed

Publication types

MeSH terms