classify

classify voxels in MRI volumes into tissue classes

classify [options] <input_volumes> <output.mnc>

DESCRIPTION

classify classifies voxels in one or more MINC input volumes into tissue classes using a variety of supervised and unsupervised classification algorithms. Supported classifiers include minimum distance, K-nearest neighbours (KNN), artificial neural network (ANN), hard C-means (HCM), fuzzy C-means (FCM), and Bayesian classification. Training data can be provided via tag files. Fuzzy classification outputs can be written as separate volumes for each tissue class.

OPTIONS

-verbose
Print progress information.
-debug
Print debugging information.
-clobber
Overwrite an existing output file.
-byte
Store output with 8-bit integer voxels.
-short
Store output with 16-bit integer voxels.
-long
Store output with 32-bit integer voxels.
-float
Store output with 32-bit floating point voxels.
-double
Store output with 64-bit floating point voxels.
-signed
Store output with signed integer voxels.
-unsigned
Store output with unsigned integer voxels.
-output_range
Specify the output intensity range.
-tagfile
Specify a tag file containing training samples.
-volume
Specify the volume index for training.
-class_names
Specify class names for the classifier.
-load_train
Load a previously saved training set.
-save_train
Save the training set for later use.
-train_only
Only perform training; do not classify.
-parameter
Specify classifier parameters.
-apriori
Specify a priori probability volumes for each class.
-fuzzy
Enable fuzzy classification output.
-fprefix
Set the prefix for fuzzy output filenames.
-fpath
Set the path for fuzzy output files.
-fuzzy_voxel_min
Set the minimum fuzzy membership for a voxel.
-fuzzy_voxel_max
Set the maximum fuzzy membership for a voxel.
-fuzzy_image_min
Set the minimum fuzzy membership for an image.
-fuzzy_image_max
Set the maximum fuzzy membership for an image.
-fbyte
Store fuzzy output with 8-bit integer voxels.
-fshort
Store fuzzy output with 16-bit integer voxels.
-flong
Store fuzzy output with 32-bit integer voxels.
-ffloat
Store fuzzy output with 32-bit floating point voxels.
-fdouble
Store fuzzy output with 64-bit floating point voxels.
-mask
Specify a mask volume for restricting classification.
-user_mask_value
Set the mask value for voxel inclusion.
-user_mask_class
Set the mask class for voxel inclusion.
-nocache
Do not use volume caching.
-max_cache_size
Set the maximum cache size in bytes.
-block_sizes
Set the block sizes for caching.
-min
Use the minimum distance classifier.
-knn
Use the K-nearest neighbours classifier.
-ann
Use the artificial neural network classifier.
-hcm
Use the hard C-means classifier.
-fcm
Use the fuzzy C-means classifier.
-bayes
Use the Bayesian classifier.

EXAMPLES

classify -tagfile training.tag t1.mnc classified.mnc

classify -knn -tagfile training.tag -mask brain_mask.mnc t1.mnc t2.mnc pd.mnc classified.mnc

classify -fcm -fuzzy -fprefix fuzzy_ t1.mnc classified.mnc

AUTHOR

Jason Lerch - McConnell Brain Imaging Centre, Montreal Neurological Institute, McGill University.

COPYRIGHTS

Copyright &copy; 2002 by Jason Lerch

SEE ALSO

nu_correct, inormalize, mincmath, volume_stats