volumes_pca

Perform principal component analysis on a set of MINC volumes.

volumes_pca [options] <training_list> <output_prefix>

DESCRIPTION

volumes_pca performs Principal Component Analysis (PCA) on a set of MINC volumes specified in a training list file. Each line of the training list contains the path to a volume. The tool computes the mean volume and the principal components of variation across the set, writing the results with the specified output prefix.

This is useful for building statistical shape or appearance models from a population of brain images.

OPTIONS

--mask mask.mnc
Restrict PCA computation to voxels within the specified binary mask.
--verbose
Print progress information during processing.
--clobber
Overwrite existing output files.
--normalize
Normalize the input volumes before PCA.
--nobias
Do not apply bias correction during PCA.
--threshold f
Set the variance threshold for retaining principal components. Components are retained until the cumulative explained variance reaches this fraction. Default: 0.98.

EXAMPLES

Compute PCA on a set of volumes:

volumes_pca training_list.txt pca_output_

Compute PCA with normalization and a mask:

volumes_pca --normalize --mask brainmask.mnc training_list.txt pca_output_

Retain 95% of variance:

volumes_pca --threshold 0.95 training_list.txt pca_output_

AUTHOR

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

COPYRIGHTS

Copyright &copy; 2011 by Vladimir S. Fonov

SEE ALSO

volumes_pca_preprocess, volumes_lm, grids_pca