Manual for ICA-AROMA
1
Introduction
ICA-AROMA (i.e. ‘ICA-based Automatic Removal Of Motion Artifacts’) attempts to identify
and remove motion artifacts from fMRI data. To that end it exploits Independent Component
Analysis (ICA) to decompose the data into a set of independent components. Subsequently,
ICA-AROMA automatically identifies which of these components are related to head motion, by
using four robust and standardized features. The identified components are then removed from
the data through linear regression as implemented in fsl_regfilt. ICA-AROMA has to be
applied after spatial smoothing, but prior to temporal filtering within the typical fMRI
preprocessing pipeline. Two manuscripts provide a detailed description and evaluation of
ICA-AROMA:
1. Pruim, R.H.R., Mennes, M., van Rooij, D., Llera, A., Buitelaar, J.K., Beckmann, C.F.,
2015, ICA-AROMA: A robust ICA-based strategy for removing motion artifacts from
fMRI data. NeuroImage
2. Pruim, R.H.R., Mennes, M., Buitelaar, J.K., Beckmann, C.F., 2015. Evaluation of
ICA-AROMA and alternative strategies for motion artifact removal in resting-state fMRI.
NeuroImage
2
General info
The ICA-AROMA package contains two python scripts; ICA_AROMA.py and
ICA_AROMA_functions.py. The first is the main script to be called by the user. The second
contains the functions used by the main script. The package furthermore contains three spatial
maps (CSF, edge & out-of-brain masks) which are required to derive the spatial features used by
ICA-AROMA. Check the ‘help’ function of ICA_AROMA.py for information on the argument
options for running ICA-AROMA.
The scripting-examples provided in this manual are based on the bash shell.
3
Requirements
• FSL
• Python 2.7 (modules: os, argparse, commands, numpy, random)
1
4
Run ICA-AROMA - generic
For standard use, ICA_AROMA.py requires the following five inputs:
-i, -in
-o, -out
-a, -affmat
-w, -warp
-mc
Input file name of fMRI data (.nii.gz)
Output directory name
File name of the mat-file describing the affine registration
(e.g. FSL FLIRT) of the functional data to structural space
(.mat file)
File name of the warp-file describing the non-linear
registration (e.g. FSL FNIRT) of the structural data to MNI152
space (.nii.gz)
File name of the text file containing the six (column-wise)
realignment parameters time-courses derived from
volume-realignment (e.g. MCFLIRT)
Example:
python2.7 ICA_AROMA.py -in func_smoothed.nii.gz -out ICA_AROMA
-affmat reg/func2highres.mat -warp reg/highres2standard_warp.nii.gz -mc
mc/rest_mcf.par
Of note, the registration files are required to transform the obtained ICA components to the
MNI152 2mm template in order to derive standardized spatial feature scores. The fMRI data
itself will not be subjected to any registration, transformation or reslicing!
4.1
Masking
Either the input fMRI data should be masked (i.e. brain-extracted) or a specific mask has to be
specified (-m, -mask) when running ICA-AROMA.
Example:
python2.7 /ICA_AROMA.py -in func_smoothed.nii.gz -out ICA_AROMA
-mc mc/rest_mcf.par -affmat reg/example_func2highres.mat -warp
reg/highres2standard_warp.nii.gz -m mask_aroma.nii.gz
We recommend not to use the mask determined by FEAT. This mask is optimized to be used for
first-level analysis, as has been dilated to ensure that all ‘active’ voxels are included. We advise
to create a mask using the Brain Extraction Tool of FSL (fractional intensity of 0.3), on a
non-brain-extracted example or mean functional image (e.g. example_func within the FEAT
directory).
Example to create an appropriate mask:
bet
Source Exif Data:
File Type : PDF
File Type Extension : pdf
MIME Type : application/pdf
PDF Version : 1.5
Linearized : No
Page Count : 4
Producer : MiKTeX pdfTeX-1.40.13
Creator : TeX
Create Date : 2015:03:25 09:13:31+01:00
Modify Date : 2015:03:25 09:13:31+01:00
Trapped : False
PTEX Fullbanner : This is MiKTeX-pdfTeX 2.9.4535 (1.40.13)