Usage
  • 7 views
  • 7 downloads

Ultrasonic heart image segmentation using active contour model

  • Author / Creator
    Sun, Jiuyu
  • This thesis is concerned with the ultrasonic heart image segmentation problem using parametric active contour model. Most of the existing parametric models consider only either the edge or the regional information. In this thesis, we propose a new parametric active contour model considering both the edge and regional information. The model is based on the vector field convolution (VFC) model and the Chan-Vese model and thus it is called as VFCCV model. The VFCCV not only has a large capture range and the ability to detect concavity, but is also robust to noise, cluttered image background, and weak boundary. Experiments on synthetic images, heart ultrasound images and MRI images show VFCCV has better performance than either VFC or Chan-Vese, and a state-of-the-art model combining edge and region information. In order to use VFCCV to segment the 3D ventricle ultrasound image, the 3D image is divided into 2D slices, and we segment each slice using VFCCV. However, the segmentation results of some slices may be inaccurate, while manual validation for every slice would be tedious. Therefore, we propose a novel automatic contour validation method, and introduce the validation step in the segmentation system. The validation step can judge whether the contour on each slice is accurate in order to make the whole process as automatic as possible while keeping a high segmentation accuracy. We extend an existing method that is based on the principal component analysis (PCA) for contour validation by performing the validation locally rather than globally in order to detect local errors. Experiments are conducted using the heart phantom ultrasound and human heart ultrasound images. Experimental results show that the VFCCV can generate correct segmentation results in most of the cases, and that the validation step can detect the errors when the contours are inaccurate. Also, the segmentation system is significantly faster than manual validation. For the heart phantom dataset, experiments show that the final ventricle volumes have high accuracy, and the system has high reproducibility.

  • Subjects / Keywords
  • Graduation date
    2014-11
  • Type of Item
    Thesis
  • Degree
    Master of Science
  • DOI
    https://doi.org/10.7939/R3MK65G0X
  • License
    This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.
  • Language
    English
  • Institution
    University of Alberta
  • Degree level
    Master's
  • Department
    • Department of Computing Science
  • Supervisor / co-supervisor and their department(s)
    • Zhang, Hong (Computing Science)
    • Ray, Nilanjan (Computing Science)
  • Examining committee members and their departments
    • Yang, Herb (Computing Science)
    • Zhang, Hong (Computing Science)
    • Ray, Nilanjan (Computing Science)