A Modified Watershed Transformation for Distorted Brain Tissue Segmentation

作者：Zanaty E A^*

来源：Journal of Medical Imaging and Health Informatics, 2012, 2(4): 393-399.

DOI：10.1166/jmihi.2012.1107

• Medical Image Segmentation
• Watershed Transform
• Multi-Degree Watershed
• Seed Region Growing
• Merging Processes

In this paper, we present an approach based on combining a modified watershed transform with seed region growing for improving brain tissue segmentation. The watershed transform is modified by adding a new penalty to partition the image into varying size levels of intensity based on image structures. This penalty makes the watershed more flexible and efficient for detecting distorted tissue from segmenting magnetic resonance images (MRIs). The region growing algorithm is used to partition the levels of intensity obtained from the proposed modified watershed into regions and adapted to work automatically on the level data to produce closed contours regions. The final image segmentation is achieved using merging techniques. For merging regions, we measure the similarity using two criteria. The first criterion finds the best seed for neighboring pixel merging, and the second applies a homogeneity threshold on similar pixels taking into account the pre-merged region. The result is a set of closed boundaries without over-segmentation. The proposed method is applied on two different datasets including distorted brain images and simulated volumetric MRIs consisting of nine classes. It is shown that the proposed method can achieve accurate results in segmenting the complex structures in medical images.

• 出版日期2012-12