em_classify

Classify voxels using expectation-maximization algorithm.

em_classify [options] <input_1> ... <input_n> <output>

DESCRIPTION

em_classify performs tissue classification on one or more input MINC volumes using the Expectation-Maximization (EM) algorithm. The EM algorithm iteratively estimates class means, standard deviations, and voxel membership probabilities to segment the input into a specified number of tissue classes.

Multiple input volumes (e.g., T1, T2, PD) can be provided for multispectral classification. Initial class parameters can be supplied via command-line options or estimated from training labels. Spatial prior probability maps can also be used to guide the classification.

OPTIONS

--verbose

Print progress information during processing.

--clobber

Overwrite existing output files.

--mask mask.mnc

Restrict classification to voxels within the specified binary mask volume.

--train train.mnc

Use the given label volume to initialize class parameters (means and standard deviations) from training data.

--means c1,c2,…,cn

Specify initial class means as a comma-separated list.

--sigma s1,…,sn

Specify initial class standard deviations as a comma-separated list.

--classes n

Set the number of tissue classes for classification.

--prob p1,p2,…

Specify a priori class probabilities as a comma-separated list. Values should sum to 1.

--iter n

Set the maximum number of EM iterations.

--priors p1.mnc,p2.mnc,…,pn.mnc

Provide spatial prior probability maps for each class as a comma-separated list of MINC files.

--save_prob base

Save posterior class probability maps using the specified base filename.

--debug

Enable debug output.

--interpolate

Use interpolation when resampling inputs.

EXAMPLES

Classify a T1 volume into 3 classes using a brain mask:

em_classify --classes 3 --mask brainmask.mnc t1.mnc classified.mnc

Classify with initial means and standard deviations:

em_classify --classes 3 --means 300,700,1000 --sigma 50,80,100 t1.mnc classified.mnc

Classify using spatial priors and save probability maps:

em_classify --classes 3 --priors csf_prior.mnc,gm_prior.mnc,wm_prior.mnc \
    --save_prob prob_ --mask brainmask.mnc t1.mnc classified.mnc

AUTHOR

Vladimir S. Fonov - McConnell Brain Imaging Centre, Montreal Neurological Institute, McGill University.

COPYRIGHTS

Copyright © 2011 by Vladimir S. Fonov

SEE ALSO

classify, mrfseg, gamixture