BRAINSCut

Back to NA-MIC Brains Collaboration

Summary

BRAINSCut is a software package for segmentation of structures using artificial neural networks. Currently this tool supports the segmentation of the following structures: brain, caudate, putamen, thalamus, hippocampus, anterior cerebellum, interior posterior cerebellum, superior posterior cerebellum, corpus medullary. Future regions will include the globus pallidus, amygdala, and nucleus accumbens. The command line uses the Slicer3 execution model framework.

Progress

1. Integration with a high dimensional registration to the atlas probability map
2. Improved thresholding of the output activation maps
3. Code added to NITRC
4. Coupled Neural network with MUSH Brain to generate a brain mask without requiring tissue classification

To Do

1. Complete integration with the FANN library
2. Link to a BSD style neural network library
3. Look at the ability to use for segmentation of cortical regions

Key Investigators

• University of Iowa: Hans Johnson, Ronald Pierson, Kent Williams, Greg Harris, Vincent Magnotta

Figures

• BRAINSCut
• 

  Brain Mask generated using BRAINSMush image and BRAINSCut overlaid on T1 for Subject 1

• 

  Brain Mask generated using BRAINSMush image and BRAINSCut overlaid on T2 for Subject 1

• 

  Brain Mask generated using BRAINSMush image and BRAINSCut overlaid on T1 for Subject 2

• 

  Brain Mask generated using BRAINSMush image and BRAINSCut overlaid on T2 for Subject 2

• 

  Brain Mask generated using BRAINSMush image and BRAINSCut overlaid on T1 for Subject 3

• 

  Brain Mask generated using BRAINSMush image and BRAINSCut overlaid on T2 for Subject 3

Usage

 BRAINSCut  [--processinformationaddress <std::string>] [--xml] [--echo]
             [--applyModel] [--trainModel] [--createVectors]
             [--generateProbability] [--trainModelStartIndex <int>]
             [--netConfiguration <std::string>] [--] [--version] [-h]

  Description: Automatic Segmentation using neural networks
 
  Author(s): Vince Magnotta, Hans Johnson, Greg Harris, Kent Williams

Where:

  --processinformationaddress <std::string>
    Address of a structure to store process information (progress, abort,
    etc.). (default: 0)

  --xml
    Produce xml description of command line arguments (default: 0)

  --echo
    Echo the command line arguments (default: 0)

  --applyModel
    apply the neural net (default: 0)

  --trainModel
    train the neural net (default: 0)

  --createVectors
    create vectors for training neural net (default: 0)

  --generateProbability
    Generate probability map (default: 0)

  --trainModelStartIndex <int>
    Starting iteration for training (default: 0)

  --netConfiguration <std::string>
    XML File defining AutoSegmentation parameters

  --,  --ignore_rest
    Ignores the rest of the labeled arguments following this flag.

  --version
    Displays version information and exits.

  -h,  --help
    Displays usage information and exits.


 

Links

• University of Iowa Department of Psychiatry
• University of Iowa MRI Center

Papers

• Powell S, Magnotta VA, Johnson H, Jammalamadaka VK, Pierson R, Andreasen NC. Registration and machine learning-based automated segmentation of subcortical and cerebellar brain structures. Neuroimage. 39(1):238-47, 2008.
• Magnotta VA, Heckel D, Andreasen NC, Cizadlo T, Corson PW, Ehrhardt JC, Yuh WT. Measurement of brain structures with artificial neural networks: two- and three-dimensional applications. Radiology. 211(3):781-90, 1999.