2013 Winter Project Week:MABS

• 

  Projects List

• 
• 

Key Investigators

• Greg Sharp, Nadya Shusharina, MGH
• Polina Golland, MIT

MABS is a program for multi-atlas based segmentation. It uses a database of multiple, segmented input images to produce a segmented result on a new image. The database is first "trained" to improve both accuracy and speed. Then, when given a new case, examples are selected from the atlas, registered to the new case, and then voting is performed to create an output segmentation.

Objective

• Create command line module for multi-atlas based segmentation
• If time allows, begin working on loadable module

Approach, Plan

• A command line module was started in Summer project week, but it is not complete
• It needs a way to specify the atlas and training directories, and specify output
• To start with, we will export as Slicer labelmap, then move to RTSTRUCT after converting to loadable

Progress

• First results!
• Segmented the brainstem using an atlas of two subjects.