Yujin Jang,M.S.1, Helen Hong, Ph.D.1, Jin Wook Chung, M.D.2
1Department of Multimedia Engineering, College of Information and Media, Seoul Women’s University, Seoul, Korea
2Department of Radiology, Seoul National University Hospital, Seoul, Korea
PurposeZZFor living donor liver transplantation, liver segmentation is difficult due to the variability of its shape across patients and similarity of the density of neighbor organs such as heart, stomach, kidney, and spleen. In this paper, we propose an automatic segmentation of the liver using multi-planar anatomy and deformable surface model in portal phase of abdominal contrast-en-hanced CT images.
MethodZZOur method is composed of four main steps. First, the optimal liver volume is extracted by positional information of pelvis and rib and by separating lungs and heart from CT images. Second, anisotropic diffusing filtering and adaptive threshold-ing are used to segment the initial liver volume. Third, morphological openthreshold-ing and connected component labelthreshold-ing are applied to multiple planes for removing neighbor organs. Finally, deformable surface model and probability summation map are performed to refine a posterior liver surface and missing left robe in previous step.
ResultsZZAll experimental datasets were acquired on ten living donors using a SIEMENS CT system. Each image had a matrix size of 512 x 512 pixels with in-plane resolutions ranging from 0.54 to 0.70 mm. The slice spacing was 2.0 mm and the number of images per scan ranged from 136 to 229. For accuracy evaluation, the average symmetric surface distance (ASD) and the volume overlap error (VE) between automatic segmentation and manual segmentation by two radiologists are calculated. The ASD was 0.26±0.12mm for manual1 versus automatic and 0.24±0.09mm for manual2 versus automatic while that of inter-radiologists was 0.23±0.05mm. The VE was 0.86±0.45% for manual1 versus automatic and 0.73±0.33% for manaual2 versus automatic while that of inter-radiologist was 0.76±0.21%.
ConclusionZZOur method can be used for the liver volumetry for the pre-surgery planning of living donor liver transplantation.
Key WordsZZLiving donor liver transplantation ㆍAbdominal CT image ㆍLiver volumetry ㆍLiver segmentation.
Received: July 11, 2014 / Revised: July 11, 2014 / Accepted: July 15, 2014 Address for correspondence: Helen Hong, Ph.D.
Department of Multimedia Engineering, College of Information and Media, Seoul Women’s University, 621 Hwarang-ro, Nowon-gu, Seoul, Korea Tel: 02-970-5756, Fax: 02-970-5981, E-mail: [email protected]
Journal of International Society for Simulation Surgery █ 2014;1(1):37-40
이 인접해 있는 영역에서 정확하게 분할하기 어렵고 최적화하는 데 수행시간이 오래 걸리는 문제점을 가지며7-10 형상정보 기반 분할기법은 주변과 밝기값이 유사한 영역에서 비교적 정확하게 분할될 수 있지만 간의 크기와 형상은 사람에 따라 다양하게 나 타나고, 특히 좌엽의 변이가 크기 때문에 훈련 집합의 형상정보 에 의해 분할 정확성의 영향을 많이 받는 한계점을 가진다.11-17
본 논문에서는 간의 밝기값과 유사한 신장, 위, 비장과 같은 주변기관을 제거하기 위하여 다단면도의 해부 정보를 사용하는 주변기관 제거 기법을 제안하고, 서로 다른 환자간의 간 형상 및 동일 환자의 각 슬라이스간 간 형상의 다양성으로 인해 발생하 는 자동 분할의 한계점을 극복하기 위하여 변형모델과 확률 누 적맵을 통해 간의 형상이 다양하더라도 견고하게 분할하는 기법 을 제안한다.
방 법
본 논문의 제안방법은 최적볼륨영역 산정, 후보 간 영역 추 출, 주변기관 제거 그리고 간 외곽선 보정의 네 단계로 이루어진 다 (Fig. 2).
첫째, 간 부위를 포함하는 최적의 간 볼륨을 추출하기 위하 여, 관상 단면도에서 밝기값 기반의 폐 및 심장 분할기법을 통해 상단경계를 정의하고 골반뼈 및 갈비뼈 인식 기법을 통해 하단경 계를 정의하며, 축상 단면도에서 갈비뼈 인식 기법을 통해 측 면 경계를 정의한다 (Fig. 3).
둘째, 산정된 최적 간 볼륨에서 간 후보영역을 추출하는 초기 간 볼륨 분할 기법을 수행하기 위하여 비등방성 발산 필터링18을 통해 간의 밝기값 정보를 갖는 영역을 강조하고 적응적으로 산 출되는 임계치 값을 통해 초기 간 볼륨을 분할한다.
셋째, 밝기값 정보를 이용한 초기 간 분할 시 간과 유사한 밝 기값으로 표현되는 주변조직과 신장, 위 그리고 비장과 같은 주 변기관이 함께 검출되는 한계점을 해결하기 위하여 다단면도에 서 해부 정보를 이용한 주변기관 제거 기법을 제안한다. 축상 단 면도에서 간과 주변 기관들은 유사한 밝기값 정보를 가지고 있 을 뿐 아니라 서로 인접하여 나타나는 해부 정보를 가지고 있기 때문에 간과 주변기관이 인접한 위치의 수직 단면도를 이용하여 주변기관을 효율적으로 제거한다. 간 영역 주변의 작은 조직 및
Fig. 1. 3The challenging issues of liver segmentation in abdomi-nal contrast-enhanced CT image : (a) Intensity similarity of neigh-bor organs such as heart, stomach, kidney and spleen (b) Vari-ability of liver intensity across patients (c) VariVari-ability of liver shape across patients.
Fig. 2. The pipeline of the proposed method for liver segmenta-tion using multi-planar anatomy and deformable surface model in abdominal contrast-enhanced CT images.
Fig. 3. The definition of optimal volume circumscribing a liver: (a) side bounding is defined in middle axial plane, (b) lower bounding is defined in middle coronal plane, and (c) upper bounding is de-fined in each coronal plane.
Fig. 4. The effect of neighbor organ elimination : (a) initial liver volume in axial plane, (b) elimination of the kidney in axial plane, (c) stomach and spleen are attached to the initial liver volume in axial plane, and (d) segmented liver after neighbor organ elimina-tion.
Fig. 5. The results of proposed method in 2D and 3D views.
Automatic Liver Segmentation on Abdominal Contrast-enhanced CT Images for the Pre-surgery Planning of Living Donor Liver Transplantation█ Hong H, et al
Fig. 6. The accuracy evaluation of segmentation results of the proposed method comparison with the manually outlining results drawn by two radiologists according to symmetric surface dis-tance and volume overlap error. (a) Average symmetric surface distance and (b) volume overlap error.
Fig. 7. Processing time per subject of proposed liver segmenta-tion method in each step: optimal volume definisegmenta-tion (2.26 sec.), initial liver volume segmentation (22.74 sec.), neighbor organs elimination (7.2 sec.) and liver border refinement (14.23 sec.).
(ax-bx)2 + (ay-by)2
Journal of International Society for Simulation Surgery █ 2014;1(1):37-40
This work was supported by Industrial Strategic technology de-velopment program, 10038419, Intelligent image diagnosis and therapy-support system funded by the Ministry of Knowledge Economy (MKE, Korea)
참 고 문 헌
1. L. Massoptier, S. Casciaro, “A new fully automatic and robust al-gorithm for fast segmentation of liver tissue and tumors form CT scans,” Eur. Radiol. vol. 18, no. 8, pp. 1658-1665, 2008
2. P. Campadelli, E. Casiraghi, A. Esposito, "Liver segmentation from computed tomography scans: a survey and a new algorithm,"
Artificial Intelligence in Medicine, vol. 45, pp. 185-196, 2009 3. A. H. Foruzan, R. A. Zoroofi, M. Hori, Y. Sato, "A
knowledge-based technique for liver segmentation in CT data," Computerized Medical Imaging and Graphics, vol. 3, no. 8, pp. 567-587, 2009 4. A. H. Foruzan, R. A. Zoroofi, Hori, Y. Sato, "Liver segmentation
by intensity analysis and anatomical information in multi-slice CT images," Int. J. CARS2009 vol.4, no. 3, pp. 287-297. 2007
5. F. Liu, B. Zhao, P. K. Kijewski, "Liver segmentation for CT images using GVF snake," Medical Physics vol. 32, no. 12, pp.3699-3707, 6. L. Rusko, G. Bekes, M. Fidrich, "Automatic segmentation of the 2005 liver from multi- and single-phase contrast-enhanced CT images,"
Medical Image Analysis, vol. 13, no. 6, pp. 871-882, 2009 7. D. Furukawa, A. Shimizu, H. Kobatake, "Automatic liver
segmen-tation method based on maximum a posterior probability estima-tion and level set method," MICCAI Workshop 3-D Segmentaestima-tion Clinic: A Grand Challenge, pp.117-124, 2007
8. E. van Rikxoort, Y. Arzhaeva, B. van Ginneken, "Automatic seg-mentation of the liver in computed tomography scans with voxel classification and atlas matching," MICCAI Workshop 3-D Seg-mentation. Clinic: A Grand Challenge, pp. 101-108, 2007
9. R. Susomboon, D. S. Raicu, J. Furst, "A hybrid Approach for Liver Segmentation," MICCAI2009, Grand Challenge, pp. 151-160, 2009 10. M. Freiman, O. Eliassaf, Y. Taieb, L. Joskowicz, Y. Azraq, J. Sos-na, "An iterative Bayesian approach for nearly automatic liver seg-mentation_algorithm and validation," Int. J. CARS, vol. 3, no. 5, pp. 439-446, 2008
11. D. Kainmuller, T. Lange, H. Lamecker, "Shape constrained auto-matic segmentation of the liver based on an heuristic intensity model," MICCAI Workshop 3-D Segmentation Clinic: A Grand Challenge, pp.109-116, 2007
12. T. Heimann, H.-P. Meinzer, and I. Wolf, “A statistical deformable model for the segmentation of liver CT volumes,” in Proc MICCAI Workshop on 3D Segmentation in the Clinic: a Grand Challenge, 2007, pp. 161–166, 2007
13. J. Liu, J. K. Udupa, "Oriented active shape models," IEEE Transac-tions on Medical Imaging, vol. 28, no. 4, pp. 571-584, 2009 14. X. Zhang, J. Tian, K. Deng, Y. Wu, X. Li, "Automatic liver
seg-mentation using a statistical shape model with optimal surface de-tection," IEEE Transactions on Biomedical Engineering, vol. 57, no. 10, pp. 2622-2626, 2010
15. A. Wimmer, G Soza, J. Hornegger, "A generic probabilistic active shape model for organ segmentation," MICCAI2009 vol. 5765, pp.
26-33, 2009
16. M. G. Linguraru, J. K. Sandberg, Z. Li, J. A. Pura, R. M. Summers,
"Atlas-based automated segmentation of spleen and liver using adaptive enhancement estimation," MICCI2009, vol. 5762, pp.1001-1008, 2009
17. M. G. Linguraru, et.al., "Automated liver segmentation using a normalized probabilistic atlas," Medical Imaging 2009, Proc. of SPIE, vol.7262, pp. 72622R-72622R-8, 2009
18. P. Perona, J. Malik : ‘Scale-space and edge detection using aniso-tropic diffusion,” IEEE Transactions on Pattern Analysis and Ma-chine Intelligence, vol. 12, no. 7, pp. 629-639, 1990
Journal of International Society for Simulation Surgery 2014;1(1):41-44
Introduction
The complete resection of a gastric tumor and adjacent lymph nodes dissection is considered the only proven, effective curative treatment option.1,2 For complete resection of the tu-mor, preoperative surgical planning is very important. Diag-noses through various imaging technology are replacing or supplementing invasive endoscopic and angiographic proce-dures.3 By using images from computed tomography (CT), not only the disease extent, but also patient-specific anatomy can be obtained before the surgical procedure.4 The introduction of image-guided procedures has allowed the application of im-age-guided surgery in accordance with the desire for minimal-ly invasive procedures.4,5 Recently, technical development of multidetector-row helical CT (MDCT) and 3-dimensional (3D) display represent significant advancement for preoperative stag-ing and CT angiography for evaluation of perigastric vascular anatomy. Surgeon could get detail information about the patient’s vascular anatomy and design of minimal invasive surgery with minimal unexpected bleed loss.6