Sub-Volume Probabilistic Atlas Segmentaiton (SVPA-SEG)

This workflow takes a raw unskull-stripped volume and a brain mask of the volume to create a tissue segmented image.This workflow is based on a novel genetic algorithm based finite mixture model and a local 3D Markov random field segmentation algorithm based on iterative conditional modes algorithm.

Problem addressed by this workflow

This workflow performs tissue segmentation on the brain volumes using genetic algorithm based finite mixture model (GAMIXTURE), local 3D Markov random field (MRF) and segmentation algorithm(SVPASEG) which is based on iterative conditional modes algorithm (ICM). The key novelty in the SVPASEG algorithm is its ability to use different MRF models on different parts of the brain.

Detailed Workflow Usage & Specifications

• Input: A raw unskull-stripped volumetric data and a binary brain mask of the volume is the input to the workflow. An atlas specification file is also input to the workflow.
• GAMIXTURE: Genetic Algorithm for mixture model optimization is used to estimate the image parameters
• SVPASEG: Voxels in the 3D image are classified into thier tissue types based on these parameters and also the atlas specification file.
• Atlas Specification Files: The Atlas specification files are used to input the information about SVPA, MRF and FMM constraints to the programs. The files that impose constraints to certain mixture components start with an identifier 'p', 'r' or 't'. P: svpaseg considers all the labels (pure and mixed) and reclassifies those voxels originally labeled as mixed voxels. r: svpaseg considers only pure labels during the tissue classification. t: svpaseg considers only pure labels during the tissue classification and uses tissue probability maps to aid the tissue classification.
• The output of this workflow is a tissue classified map for each volume. Each of the tissue types are weighted with different value. For the 3 tissue type classification, 0: background; 1: CSF (Cerebro Spinal Fluid); 2: GM (Gray Matter); 3: WM (White Matter).

URL: http://www.loni.ucla.edu/twiki/bin/view/CCB/PipelineWorkflows_SVPASEG

Try this Pipeline Workflow Now!