Name: FeenICS
Owner: Kimel Family Translational Imaging-Genetics Research Lab
Description: A pipeline for FrEquENcy-based Ica Cleaning of Spirals
Created: 2017-11-21 20:52:35.0
Updated: 2017-12-29 19:30:34.0
Pushed: 2018-01-15 19:52:59.0
Homepage: null
Size: 1024
Language: Python
GitHub Committers
| User | Most Recent Commit | # Commits |
|---|
Other Committers
| User | Most Recent Commit | # Commits |
|---|
A pipeline for FrEquENcy-based Ica Cleaning of Spirals. Aims to remove spiral-specific aritfact from fMRI scans.
This tool was built to work with FSL's MELODIC. It was created to automate the process of evaluating and removing ICA components containing spiral artifact, which is accomplished by evaulating the per slice frequency loading, and scoring components based on patterns of low or high frequency information. The fsl_regfilt command-line program can then be used to regress out components selected for removal.
ownload Dockerfile
udo apt-get install docker.io (if you don't have docker already installed)
d /path/to/Dockerfile
udo docker build -t feenics .
udo docker run -v /path/to/data:/input -i -t feenics
NOTE: where /path/to/data is a folder that you want access to from inside
the container; 'input' will be the name of this folder inside the container.
This program requires either FSL or the outputs of FSL Melodic in order to work.
Python dependancies (these will be installed when you install FeenICS if you don't already have them):
e:
_folder_setup.py -p -i PATH -s PATH <directory>
ments:
irectory> Path to top experiment directory. Outputs will be created
here.
ons:
sprl, -i PATH Specify alternative folder containing split spirals.
subs, -s PATH Specify alternative experiment folder containing
subjectID subfolders.
parallel, -p Print instuctions for running in parallel (GNU parallel)
instead of running FSL steps on machine in series.
ILS
s subject and sprl subfolders within <directory>. Moves separated spiral
s to appropriate subfolders. If split spirals are not already contained
in <directory>, use the "-i" option to specify an alternative path. Specify
if you would not like to run FSL preprocessing steps at this time. It will
t instructions to run MCFLIRT, BET, and MELODIC using GNU parallel.
e:
_identify_components.py -m FLOAT -l FLOAT <directory>
ments:
irectory> Path to top experiment directory.
ons:
midFactor, -m Cutoff multiplier for mid range frequency information.
Raise this value to more aggressively remove noise
components. Default is 3.
lowFactor, -l Cutoff multiplier for low range frequency information.
Raise this value to keep more signal components.
Default is 1.
ILS
s paths, multipliers, and output file names to check_slices.py for each
ect within the specified directory. The default name for the output
sification file is fix4melview_Standard_thr20.txt.
e:
_remove_flagged_components.py -c PATH -o PATH <directory>
ments:
irectory> Path to top experiment directory.
ons:
clean_img, -c Path to desired location of cleaned images.
output, -o If csv outputs were generated in non-default location,
identify path to this location.
ILS
fsl_regfilt to regress out the components specified in the last line of the
sification file (fix4melview_Standard_thr20.txt). The outputs are cleaned,
still separated spiral niftis. They will be named subject.sprl.denoised.nii.gz.
An html report creation tool written by E.Dickie. This step is optional, but will make checking your data much easier.
Follow the sample in Usage Examples, below or click here for futher instructions.
e:
eck_slices.py <melodic_file> <outputname> <factorA> <factorB>
ments:
elodic_file> Path to any melodic_IC.nii.gz file.
utputname> Path to desired output classification file location.
actorA> Multiplier to be used to determine mid/high frequency
cutoffs. If called from s2, default is 3.
actorB> Multiplier to be used to determine low frequency cutoffs.
If called from s2, default is 1.
ILS
script is called by s2_identify_components.py. Can be used independently to
bleshoot classification or path identification issues, or just to run one
al scan at a time. The algorithm scores each slice based on how much high
low frequency information is present after performing an FFT (fast fourier
sform). The scores are tallied per component, and the result determines
her or not a component is to be flagged for removal. The output of this script
classification text file (fix4melview_Standard_thr20.txt), which details the
e and classification criteria per slice, as well as the removal decision
component. The final line of this document is a list of which components are
e removed. If you disagree with the decision, change the component numbers in
list, as this line will be read into s3_remove_flagged_components.py.
To run FeenICS locally for an experiment called EXPR, with additional use of Erin Dickie's ICArus package (https://github.com/edickie/ICArus) to better visualize the decisions made by the check_slices.py algorithm
older_setup.py /path/to/EXPR
dentify_components.py /path/to/EXPR
s=`cd /path/to/EXPR; ls -1d */sprl*/`
us-report ${sprls}
qc_icafix.html in your favourite web browser to view the classification results.
emove_flagged_components.py /path/to/EXPR