https://github.com/dangom/tfm

Compute temporal functional modes (or atlas based temporal ICA) from functional MRI data.
https://github.com/dangom/tfm

Last synced: about 2 months ago
JSON representation

Compute temporal functional modes (or atlas based temporal ICA) from functional MRI data.

• Host: GitHub
• URL: https://github.com/dangom/tfm
• Owner: dangom
• Created: 2018-09-17T19:10:20.000Z (over 6 years ago)
• Default Branch: master
• Last Pushed: 2019-05-15T22:42:32.000Z (over 5 years ago)
• Last Synced: 2023-10-05T20:31:31.626Z (over 1 year ago)
• Language: Python
• Size: 445 KB
• Stars: 3
• Watchers: 5
• Forks: 3
• Open Issues: 2
• Metadata Files:
  • Readme: README.org

Awesome Lists containing this project

README

        
#+TITLE: Temporal Functional Modes

Code to compute Temporal Functional Modes.

In other words, run a temporal ICA on the timeseries of spatial ICA, and then remix the spatial maps to generate 'modes' of brain activity. For more information take a look at:

Temporally-independent functional modes of spontaneous brain activity (Smith et al. 2012).

Or done at the single-subject level at:

[[http://index.mirasmart.com/ISMRM2019/PDFfiles/0911.html][Subject specific functional connectivity fingerprints made possible with temporal ICA]]

Temporal Functional Modes as described in Smith 2012 are actually nothing but temporal ICA run on parcels of the brain, instead of voxels directly. Provided there are enough timepoints for tICA to converge, there is little difference in running the TFM algorithm directly on signals obtained from an atlas based parcellation, instead of a spatial ICA parcellation (as suggested in the original TFM contribution). Therefore, this repository also ships with the MIST parcellation atlas for directly application of temporal ICA for fMRI data.

* Installation

TFM only supports python 3.6+ (because of f-strings). Either clone the project and install it by running

#+BEGIN_SRC sh

    $ python setup.py install

#+END_SRC

or install it directly with pip by calling:

#+BEGIN_SRC sh

    $ pip install git+https://github.com/dangom/tfm.git

#+END_SRC

* Usage

TFM provides a command line tool called =tfm=, which allows one to configure the tICA algorithm, as well as inputs and outputs. In essence, users can pass a demeaned and denoised fMRI dataset directly as an input, in which case an atlas based parcellation will take place (atlas mode). Otherwise, passing an FSL MELODIC output directory (with melodic_mix, melodic_oIC.nii.gz and melodic_unmix files) will compute TFMs as described in Smith 2012 (ICA mode). Users can also pass a dual_regression directory, so that the =dr_stage1_subject0000.txt= file with timeseries will be used as signal timeseries, and the =dr_stage2_subject0000.nii.gz=  maps will be used as RSN maps (Dual Regression Mode).

Output results of TFM are saved with names following the MELODIC convention, only so as to allow usage and visualization using FSLEYES.

* Outputs

The TFM routine outputs the following files upon completion:

| Filename                            | Meaning                                  | Created when...           |

|-------------------------------------+------------------------------------------+---------------------------|

| melodic_IC.nii.gz                   | TFM maps                                 | always                    |

| melodic_unmix                       | TFM core mixing matrix                   | always                    |

| melodic_mix                         | TFM timeseries                           | always                    |

| melodic_FTmix                       | TFM frequency spectra                    | always                    |

| tfm.log                             | TFM routine log                          | always                    |

| network_contributions.csv           | 12-network summary of melodic_unmix      | atlas mode                |

| network_contributions_64.csv        | 64-network summary of melodic_unmix      | atlas mode                |

| melodic_unmix.png                   | Plot of network_contributions.csv        | atlas mode                |

| tfm_correlation_to_confounds.csv    | TFM timeseries correlation to cof.       | presence of confound file |

| signal_correlation_to_confounds.csv | Signal timeseries correlation to cof.    | presence of confound file |

| correlation_with_confounds.png      | Plot of both signal and tfm corr to cof. | presence of confound file |

* Disclaimer

Some features of TFM (such an experimental raicar and isctest submodules) are under development, considered experimental or not tested at all. Use at your own risk.

* License

tfm is distributed under the terms of the

- [[https://choosealicense.com/licenses/gpl-3.0][GNU GPL v3]]