2013 Summer Project Week:Multi Atlas Based Multi Image Segmentation
Key Investigators
- UNC: Minjeong Kim, Dinggang Shen
- GE: Xiaofeng Liu, Jim Miller
Objective
The accuracy of existing multi-atlas-based segmentation methods relies heavily on the precise alignment of the atlases with the target image. Moreover, when segmenting a group of target images, these images are considered independently with disregard of their correlation, thus resulting in inconsistent segmentations of the same structures across different target images. To address these issues, we have developed a method for iterative multi-atlas-based multi-image segmentation with tree-based registration and will extend it to be used in Slicer 4.
Approach, Plan
Our method consists of 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.
Progress
We have implemented a command-line version of Slicer module. Testing with various datasets is ongoing. Meanwhile, we will also develop GUI version of Slicer module.
Delivery Mechanism
This work will be delivered to the NA-MIC Kit as a Slicer module (extension).
References
- Hongjun Jia, Pew-Thian Yap, Dinggang Shen, "Iterative multi-atlas-based multi-image segmentation with tree-based registration", NeuroImage 59, pp,422-430 (2012).