Elsevier

Academic Radiology

Volume 15, Issue 11, November 2008, Pages 1390-1403
Academic Radiology

Medical image computing and computer-assisted intervention
Automated Segmentation of the Liver from 3D CT Images Using Probabilistic Atlas and Multilevel Statistical Shape Model

https://doi.org/10.1016/j.acra.2008.07.008Get rights and content

Rationale and Objectives

An atlas-based automated liver segmentation method from three-dimensional computed tomographic (3D CT) images has been developed. The method uses two types of atlases, a probabilistic atlas (PA) and a statistical shape model (SSM).

Materials and Methods

Voxel-based segmentation with a PA is first performed to obtain a liver region, then the obtained region is used as the initial region for subsequent SSM fitting to 3D CT images. To improve reconstruction accuracy, particularly for highly deformed livers, we use a multilevel SSM (ML-SSM). In ML-SSM, the entire shape is divided into patches, with principal component analysis applied to each patch. To avoid inconsistency among patches, we introduce a new constraint called the “adhesiveness constraint” for overlapping regions among patches.

Results

The PA and ML-SSM were constructed from 20 training datasets. We applied the proposed method to eight evaluation datasets. On average, volumetric overlap of 89.2 ± 1.4% and average distance of 1.36 ± 0.19 mm were obtained.

Conclusions

The proposed method was shown to improve segmentation accuracy for datasets including highly deformed livers. We demonstrated that segmentation accuracy is improved using the initial region obtained with PA and the introduced constraint for ML-SSM.

Section snippets

Spatial Normalization Using the Abdominal Cavity

Given training datasets, PA and ML-SSM are constructed. Before construction, spatial normalization of datasets is necessary. To achieve this, one CT dataset that was judged to display average liver shape by a radiology specialist was selected as a standard patient, and the abdominal cavity was regarded as the normalized space to represent normalized liver position and shape (12). A rough abdominal cavity is used in this work, because extracting the precise region of the abdominal cavity is not

Experimental Conditions

A total of 28 abdominal CT datasets (slice thickness 2.5 mm, pitch 1.25 mm, field of view 350 × 350 mm2, 512 × 512 matrix, 159 slices) were used. We randomly selected eight datasets for evaluation, with the others used for training. The PA and ML-SSM were constructed from the 20 training datasets. A radiology specialist judged the livers as highly deformed from disease in nine of the 20 training datasets. Livers were also highly deformed in five of the eight evaluation datasets.

In the nonrigid

Discussion

In cases without the adhesiveness constraint, when the hierarchy of ML-SSM changed from level 2 to level 3, segmentation accuracy decreased (as shown in Table 1) and large inconsistencies among patches of the estimated shape occurred. When the hierarchy of ML-SSM changed from level 1 to level 2, large inconsistencies occurred although segmentation accuracy increased. For large λ (λ ≥ 0.05), segmentation accuracy was improved as the hierarchy level of ML-SSM increased, and inconsistency could be

Conclusion

We have developed an automated segmentation method of the liver using statistical atlases. The proposed method was shown to improve segmentation accuracy for datasets including highly deformed livers by combining initial segmentation based on PA and subsequent ML-SSM fitting. We also demonstrate that the adhesiveness constraint was effective for dealing with inconsistencies at finer levels of ML-SSM. Accuracy was degraded at the fine level (level 3) without use of the constraint. We are now

References (20)

  • T.F. Cootes et al.

    Active shape models—their training and application

    Comp Vision Image Understanding

    (1995)
  • S. Joshi et al.

    Unbiased diffeomorphic atlas construction for computational anatomy

    NeuroImage

    (2004)
  • H. Chui et al.

    A new point matching algorithm for non-rigid registration

    Computer Visi Image Understanding

    (2003)
  • M.E. Leventon et al.

    Statistical shape influence in geodesic active contours

    Proc IEEE Computer Soc Conf Computer Vision Pattern Recognition

    (2000)
  • J. Bailleul et al.

    Statistical shape model-based segmentation of brain MRI images

  • D. Shen et al.

    An adaptive-focus statistical shape model for segmentation and shape modeling of 3-D brain

    IEEE Trans Med Imaging

    (2001)
  • H. Park et al.

    Construction of an abdominal probabilistic atlas and its application in segmentation

    IEEE Trans Med Imaging

    (2003)
  • M. Straka et al.

    Bone segmentation in CT-angiography data using a probabilistic atlas

    Vision Modeling Visualization

    (2003)
  • H. Lamecker et al.

    Segmentation of the liver using a 3D statistical shape model

    (2004)
  • T. Heimann et al.

    Active shape models for a fully automated 3D segmentation of the liver— an evaluation on clinical data

    (2006)
There are more references available in the full text version of this article.

Cited by (114)

  • Shape prior model via dual subspace segment projection learning

    2021, Computer Methods and Programs in Biomedicine
View all citing articles on Scopus
View full text