2014 Project Week:MultiAtlas MultiImage Segmentation
Key Investigators
- Minjeong Kim, Dinggang Shen, UNC Chapel Hill
- Xiaofeng Liu, Jim Miller, GE Research
Project Description
Objective
We develop a Slicer module for multi-atlas-based multi-image segmentation of brain images. To deal with the limitation of existing pairwise registration methods between images with large shape difference, our algorithm in the module performs 1) a novel tree-based groupwise registration method for concurrent alignment of both the atlases and the target images, and 2) an iterative groupwise segmentation method for simultaneous consideration of segmentation information propagated from all available images, including the atlases and other newly segmented target images.
Approach, Plan
We have developed the Slicer module called MABMIS and fully tested it using LONI LPBA40 and IXI datasets. The result by our Slicer module shows 2% improvement compared to pairwise registration methods in terms of the averaged overlap ratio between automatic segmentation and ground truth labels.
Progress
We are releasing our Slice module in NITRC (https://www.nitrc.org).
