Journal article
Multilevel segmentation of intracranial aneurysms in CT angiography images.
- Wang Y Radiology and Biomedical Imaging, University of California, San Francisco, San Francisco, California 94122 and University of Lyon, CREATIS, CNRS UMR 5220, INSERM U1206, UCB Lyon1, INSA Lyon, Lyon 69100, France.
- Zhang Y Veterans Affairs Medical Center, San Francisco, California 94121 and University of Lyon, CREATIS, CNRS UMR 5220, INSERM U1206, UCB Lyon1, INSA Lyon, Lyon 69100, France.
- Navarro L Ecole Nationale Superieure des Mines de Saint-Etienne, Saint-Etienne 42015, France.
- Eker OF CHU Montpellier, Neuroradiologie, Montpellier 34000, France.
- Corredor Jerez RA Ecole Polytechnique Federale de Lausanne, Lausanne 1015, Switzerland.
- Chen Y University of Lyon, CREATIS, CNRS UMR 5220, INSERM U1206, UCB Lyon1, INSA Lyon, Lyon 69100, France.
- Zhu Y University of Lyon, CREATIS, CNRS UMR 5220, INSERM U1206, UCB Lyon1, INSA Lyon, Lyon 69100, France.
- Courbebaisse G University of Lyon, CREATIS, CNRS UMR 5220, INSERM U1206, UCB Lyon1, INSA Lyon, Lyon 69100, France.
- 2016-04-03
Published in:
- Medical physics. - 2016
Adult
Aged
Computed Tomography Angiography
Female
Humans
Image Processing, Computer-Assisted
Intracranial Aneurysm
Male
Middle Aged
English
PURPOSE
Segmentation of aneurysms plays an important role in interventional planning. Yet, the segmentation of both the lumen and the thrombus of an intracranial aneurysm in computed tomography angiography (CTA) remains a challenge. This paper proposes a multilevel segmentation methodology for efficiently segmenting intracranial aneurysms in CTA images.
METHODS
The proposed methodology first uses the lattice Boltzmann method (LBM) to extract the lumen part directly from the original image. Then, the LBM is applied again on an intermediate image whose lumen part is filled by the mean gray-level value outside the lumen, to yield an image region containing part of the aneurysm boundary. After that, an expanding disk is introduced to estimate the complete contour of the aneurysm. Finally, the contour detected is used as the initial contour of the level set with ellipse to refine the aneurysm.
RESULTS
The results obtained on 11 patients from different hospitals showed that the proposed segmentation was comparable with manual segmentation, and that quantitatively, the average segmentation matching factor (SMF) reached 86.99%, demonstrating good segmentation accuracy. Chan-Vese method, Sen's model, and Luca's model were used to compare the proposed method and their average SMF values were 39.98%, 40.76%, and 77.11%, respectively.
CONCLUSIONS
The authors have presented a multilevel segmentation method based on the LBM and level set with ellipse for accurate segmentation of intracranial aneurysms. Compared to three existing methods, for all eleven patients, the proposed method can successfully segment the lumen with the highest SMF values for nine patients and second highest SMF values for the two. It also segments the entire aneurysm with the highest SMF values for ten patients and second highest SMF value for the one. This makes it potential for clinical assessment of the volume and aspect ratio of the intracranial aneurysms.
Segmentation of aneurysms plays an important role in interventional planning. Yet, the segmentation of both the lumen and the thrombus of an intracranial aneurysm in computed tomography angiography (CTA) remains a challenge. This paper proposes a multilevel segmentation methodology for efficiently segmenting intracranial aneurysms in CTA images.
METHODS
The proposed methodology first uses the lattice Boltzmann method (LBM) to extract the lumen part directly from the original image. Then, the LBM is applied again on an intermediate image whose lumen part is filled by the mean gray-level value outside the lumen, to yield an image region containing part of the aneurysm boundary. After that, an expanding disk is introduced to estimate the complete contour of the aneurysm. Finally, the contour detected is used as the initial contour of the level set with ellipse to refine the aneurysm.
RESULTS
The results obtained on 11 patients from different hospitals showed that the proposed segmentation was comparable with manual segmentation, and that quantitatively, the average segmentation matching factor (SMF) reached 86.99%, demonstrating good segmentation accuracy. Chan-Vese method, Sen's model, and Luca's model were used to compare the proposed method and their average SMF values were 39.98%, 40.76%, and 77.11%, respectively.
CONCLUSIONS
The authors have presented a multilevel segmentation method based on the LBM and level set with ellipse for accurate segmentation of intracranial aneurysms. Compared to three existing methods, for all eleven patients, the proposed method can successfully segment the lumen with the highest SMF values for nine patients and second highest SMF values for the two. It also segments the entire aneurysm with the highest SMF values for ten patients and second highest SMF value for the one. This makes it potential for clinical assessment of the volume and aspect ratio of the intracranial aneurysms.
- Language
-
- English
- Open access status
- closed
- Identifiers
-
- DOI 10.1118/1.4943375
- PMID 27036575
- Persistent URL
- https://sonar.ch/global/documents/96984
Statistics
Document views: 46
File downloads: