Labeling of white matter of MRI brain images using skeleton

  • Author / Creator
    Taghaddosi, Mohsen
  • Image segmentation is the problem of assigning 2D pixels or 3D voxels to a set of finite labels. In particular, in medical imaging, the goal of segmentation is to partition an MRI or CT image into regions that are relevant to the biology of a particular disease, for example the motor cortex in Amyotrophic Laterals Sclerosis (ALS). This thesis studies the problem of segmenting T1 MR images of a human brain into a set of predefined regions. Heretofore, many methods have been proposed for brain segmentation into anatomical structures, some of which use a set of manually labeled brain images to guide the segmentation. These methods use the intensity of each voxel to calculate the best transformation which registers images into a single common space and then they apply the same transformation to the labeled images to get the final results. Our work is based on extracting the skeleton (medial axis) of white matter and use it to choose an image from a set of images in a training data set that are manually labeled by experts. Then we use that information to register the input image and then apply the same transformation to the labeled image to achieve an estimation of the final result. Finally, a refining optimization step is done by applying an optimization to improve the results. Our experiments on a data set shows that our proposed method outperforms current methods in terms of accuracy and also its time complexity is very efficient.

  • Subjects / Keywords
  • Graduation date
    Spring 2016
  • Type of Item
  • Degree
    Master of Science
  • DOI
  • 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.