tedana

Analysis of echo-time dependent fMRI data

Outline

Introduction to multi-echo fMRI
Multi-echo fMRI for noise removal
tedana: the software
tedana: the community
Challenges developing tedana
Contributing to tedana

Single-echo fMRI

What is multi-echo fMRI

Signal decay

Assuming a mono-exponential decay model:

$s(x, t, TE_k) = S_0(x, t) e^{-R_2^*(x,t) \cdot TE_k}$

Describing $S_0(t)$ and $R_2^*(t)$ in terms of relative changes w.r.t. their average values:

$s(t, TE_k) = \bar{s}(TE_k) \cdot (1 + \frac{{\Delta}{S_0}(t)}{\bar{S}_0})\cdot e^{- {\Delta}{R_2^*}(t) \cdot TE_k}$

Signal percentage changes w.r.t. the mean of the signal:

$y(t, TE_k) = {{\Delta}{\rho(t)}} -{\Delta}{R_2^*(t)}\cdot TE_k$

Separating $T_2^*$ and $S_0$

Most common denoising approaches for multi-echo fMRI

Optimal combination of echoes ¹ $^,$ ²

Weighted average of echoes based on voxel-wise $T_2^*$ estimates
Reduces thermal noise and weights towards $T_2^*$ signal
Straightforward math & reliable improvements
Automatically calculated in AFNI, fMRIPrep & tedana

Denoising to actively separate $T_2^*$ and $S_0$

Based on ICA: ME-ICA ¹ $^,$ ² and tedana ³
Multi-Echo Paradigm Free Mapping ⁴ (available in AFNI and pySPFM)
Other methods

Denoising with ICA

Fit signal decay
Optimal combination of echoes
PCA to remove thermal noise
Decomposition into independent components (ICA)
Classification of components into BOLD-like (accepted) and non BOLD-like (rejected)
Denoising

Benefits of denoising with ICA

tedana: the software

Elizabeth DuPre started tedana in May 2018 to advance ME-ICA by Prantik Kundu

Approx. 20 contributors from various countries and institutions

Monthly developer calls, a periodic newsletter, active issue board & code updates

tedana: the software

Install:

pip install tedana

Use from the command line:

tedana -d echo_1.nii.gz echo_2.nii.gz echo_3.nii.gz echo_4.nii.gz -e 12 28 44 60 -

Or from a Python session:

from tedana import workflows
workflows.tedana_workflow(
    data_files,
    echo_times,
    out_dir=out_dir,
    mask=mask_file,
    prefix="sub-04570_task-rest_space-scanner",
    fittype="curvefit",
    tedpca="mdl",
    verbose=True,
)

tedana: the software

Check out the Juggling with data session on Thursday at 12:00 UTC for more on these.

tedana: the community

tedana
maPCA
Rica
Multi-echo Jupyter Book
Curated lists of multi-echo posters from OHBM
Other works-in-progress

tedana: the community

tedana
maPCA
Rica
Multi-echo Jupyter Book
Curated lists of multi-echo posters from OHBM
Other works-in-progress

tedana: the community

tedana
maPCA
Rica
Multi-echo Jupyter Book
Curated lists of multi-echo posters from OHBM
Other works-in-progress

tedana: the community

tedana
maPCA
Rica
Multi-echo Jupyter Book
Curated lists of multi-echo posters from OHBM
Other works-in-progress

tedana: the community

tedana
maPCA
Rica
Multi-echo Jupyter Book
Curated lists of multi-echo posters from OHBM
Other works-in-progress

tedana: the community

tedana
maPCA
Rica
Multi-echo Jupyter Book
Curated lists of multi-echo posters from OHBM
Other works-in-progress

Challenges developing tedana

Expertise required: software engineering, MRI physics, math/statistics, neuroscience, data visualization, project management, community management…

👉 No one in the team has all of these!

Challenges developing tedana

Neuroimaging grad students and postdocs are rarely software engineers, but their code needs to survive beyond their period of contribution

👉 Documentation and testing are very important

Challenges developing tedana

Too big for informal governance, but too small for most documented open source governance models

Leadership and Governance

Improvements prompted & guided by Kirstie Whitaker
Acknowledge contributors’ leadership & minimize administrative burden
Grow community, strive for consensus, have a path for decision making
Support openness in both code and decision making
A steering committee for project-wide steering, and focused roles for code-focused tasks
Code of conduct and enforcement policy (luckily, never used)

Contributing to tedana

Code contributions
New ideas for data processing and visualization
Documentation & education
Questions that can help us identify gaps in the code or documentation
Sharing (or processing your own) datasets to help with validation and testing
Join monthly call!

Discussions are open

Thursday at 12:00 UTC

Learn how to denoise your multi-echo fMRI data using tedana at the Juggling with data session at MRI Together

Thank you!

Get in touch @eurunuela or e.urunuela@bcbl.eu

tedana

Analysis of echo-time dependent fMRI data

Outline

Single-echo fMRI

What is multi-echo fMRI

Signal decay

Separating T2∗T_2^*T2∗​ and S0S_0S0​

Most common denoising approaches for multi-echo fMRI

Optimal combination of echoes 1,^,,2

Denoising to actively separate T2∗T_2^*T2∗​ and S0S_0S0​

Denoising with ICA

Benefits of denoising with ICA

tedana: the software

tedana: the software

tedana: the software

tedana: the community

tedana: the community

tedana: the community

tedana: the community

tedana: the community

tedana: the community

Challenges developing tedana

Challenges developing tedana

Challenges developing tedana

Leadership and Governance

Contributing to tedana

Discussions are open

Thursday at 12:00 UTC

Thank you!

Separating $T_2^*$ and $S_0$

Optimal combination of echoes ¹ $^,$ ²

Denoising to actively separate $T_2^*$ and $S_0$