interfaces.fsl.fix

Classifier

Link to code

Wraps command None

Classify ICA components using a specific training dataset (<thresh> is in the range 0-100, typically 5-20).

Inputs:

[Mandatory]
thresh: (an integer (int or long))
        Threshold for cleanup.
        flag: %d, position: -1
trained_wts_file: (an existing file name)
        trained-weights file
        flag: %s, position: 2

[Optional]
args: (a unicode string)
        Additional parameters to the command
        flag: %s
artifacts_list_file: (a file name)
        Text file listing which ICs are artifacts; can be the output from
        classification or can be created manually
environ: (a dictionary with keys which are a bytes or None or a value
         of class 'str' and with values which are a bytes or None or a value
         of class 'str', nipype default value: {})
        Environment variables
mel_ica: (an existing directory name)
        Melodic output directory or directories
        flag: %s, position: 1

Outputs:

artifacts_list_file: (a file name)
        Text file listing which ICs are artifacts; can be the output from
        classification or can be created manually

Cleaner

Link to code

Wraps command None

Extract features (for later training and/or classifying)

Inputs:

[Mandatory]
artifacts_list_file: (an existing file name)
        Text file listing which ICs are artifacts; can be the output from
        classification or can be created manually
        flag: %s, position: 1

[Optional]
aggressive: (a boolean)
        Apply aggressive (full variance) cleanup, instead of the default
        less-aggressive (unique variance) cleanup.
        flag: -A, position: 3
args: (a unicode string)
        Additional parameters to the command
        flag: %s
cleanup_motion: (a boolean)
        cleanup motion confounds, looks for design.fsf for highpass filter
        cut-off
        flag: -m, position: 2
confound_file: (a file name)
        Include additional confound file.
        flag: -x %s, position: 4
confound_file_1: (a file name)
        Include additional confound file.
        flag: -x %s, position: 5
confound_file_2: (a file name)
        Include additional confound file.
        flag: -x %s, position: 6
environ: (a dictionary with keys which are a bytes or None or a value
         of class 'str' and with values which are a bytes or None or a value
         of class 'str', nipype default value: {})
        Environment variables
highpass: (a float, nipype default value: 100)
        cleanup motion confounds
        flag: -m -h %f, position: 2

Outputs:

cleaned_functional_file: (an existing file name)
        Cleaned session data

FeatureExtractor

Link to code

Wraps command None

Extract features (for later training and/or classifying)

Inputs:

[Mandatory]

[Optional]
args: (a unicode string)
        Additional parameters to the command
        flag: %s
environ: (a dictionary with keys which are a bytes or None or a value
         of class 'str' and with values which are a bytes or None or a value
         of class 'str', nipype default value: {})
        Environment variables
mel_ica: (an existing directory name)
        Melodic output directory or directories
        flag: %s, position: -1

Outputs:

mel_ica: (an existing directory name)
        Melodic output directory or directories
        flag: %s, position: -1

Training

Link to code

Wraps command None

Train the classifier based on your own FEAT/MELODIC output directory.

Inputs:

[Mandatory]

[Optional]
args: (a unicode string)
        Additional parameters to the command
        flag: %s
environ: (a dictionary with keys which are a bytes or None or a value
         of class 'str' and with values which are a bytes or None or a value
         of class 'str', nipype default value: {})
        Environment variables
loo: (a boolean)
        full leave-one-out test with classifier training
        flag: -l, position: 2
mel_icas: (a list of items which are an existing directory name)
        Melodic output directories
        flag: %s, position: -1
trained_wts_filestem: (a unicode string)
        trained-weights filestem, used for trained_wts_file and output
        directories
        flag: %s, position: 1

Outputs:

trained_wts_file: (an existing file name)
        Trained-weights file

TrainingSetCreator

Link to code

Goes through set of provided melodic output directories, to find all the ones that have a hand_labels_noise.txt file in them.

This is outsourced as a separate class, so that the pipeline is rerun everytime a handlabeled file has been changed, or a new one created.

Inputs:

[Mandatory]

[Optional]
mel_icas_in: (a list of items which are an existing directory name)
        Melodic output directories
        flag: %s, position: -1

Outputs:

mel_icas_out: (a list of items which are an existing directory name)
        Hand labels for noise vs signal
        flag: %s, position: -1