Research > Software > MRFSEG

MRFSEG+GAMIXTURE Software Bundle

MRFSEG+GAMIXTURE is a collection of tools implementing a flexible voxel classification framework. The framework is based on a novel genetic algorithm based finite mixture model (GAMIXTURE) and a standard 3-D Markov random field (MRF) based on the iterative conditional modes (ICM) algorithm (MRFSEG).



Download MRFSEG

Select operating system *

Linux - Version 1.1 (beta)

By checking "I agree" and requesting to download software you agree that you will abide by our Software License.


Features

  • MRFSEG is programmed to automatically classify voxel into tissue types in 3-D images while modeling the spatial interactions of the voxel labels by an MRF.
  • GAMIXTURE is a program that implements a new genetic algorithm based finite mixture model (FMM) parameter estimation scheme. This allows for robustly finding parameters for an almost arbitrary 1-d FMM. However, at the moment Gaussian densities are used to model the pure tissue types.

Description

MRFSEG+GAMIXTURE is a collection of open-source software tools implementing a flexible voxel classification framework. The framework is based on a novel genetic algorithm based finite mixture model (GAMIXTURE) and a standard 3-D Markov random field (MRF) based on the iterative conditional modes (ICM) algorithm (MRFSEG). The key novelty in the MRFSEG program is the application of the configuration file to control MRF parameters. The software package is to be extended with a novel inhomogeneous MRF algorithm, allowing even more flexible spatial and intensity information modeling. At the moment the software supports Analyze 7.5 images. Future releases will add support for MINC files as well.

There are two applications in the package:

MRFSEG for brain image segmentation using Markov random fields.

GAMIXTURE for fitting FMMs to the image data using genetic algorithms.

The purpose is to 1) estimate the image parameters by GAMIXTURE, and 2) use these parameters to classify voxels in a 3-D image by MRFSEG. Ready-made shell scripts implementing this pipeline are available. These two programs together offer a very flexible framework to perform tissue classification. For example, the number of tissue classes, the MRF parameters, constraints on the FMM, and how the partial volume effect is handled can be tuned without re-compiling.

Usage

<br />

System Requirements

Linux

  • Version: 1.1 (beta)
  • Size: 141 Kb
  • OS: tested using Linux kernels 2.4 and 2.6.
  • Processor: n/a
  • Memory: The memory requirements depend on the size of the image to be tissue classified and the number of ti
  • Software: g++ compiler and Makefile utility

Installation

  1. Download the source
  2. Unpack the source
  3. cd to src sub-directory
  4. type make

Purpose

This is a command line application without graphical interface. The Analyze 7.5 image format is supported. In the near future, support for MINC is likely to be added. See the user\'s manual for more details.