Component selection decision trees in tedana

Component selection methods (decision trees) used by tedana software for multi-echo fMRI denoising. While the publication below should be cited by people who use tedana, this item can be used as a citable DOI so that users can reference the exact decision tree steps they used to process their data.

The decision trees here include:

• tedana_orig, which is the version used in tedana for years. This was called kundu in version 23, and was the unnamed and only option in earlier versions
• meica was added in version 24.0.0. The tedana developers noticed a small difference between the decision tree implemented in MEICA v2.5 which we intended to implement in tedana and what was actually implemented. The results from tedana_orig and meica will either be identical or meica will accept additional components. The additionally accepted components can have substantive variance and, upon visual inspection usually looked like they should have been rejected. Tedana's default remains tedana_orig, but meica was added to allow for direct comparisions to the older MEICA software.
• minimal is a decision tree that was released with version 23. It has fewer steps that the other trees and is closer to simpler decision criteria described in the original MEICA manuscript.
• demo_external_regressors_single_model and demo_external-regressors_motion_task_models are two decision trees that demonstrate functional for how to include external regressors, like breathing, heart rate, and head motion, into a decision tree. They are plausibly useful as is, but work to define and validate best practices is still in progress
• robust_tedana_2025 is based on tedana_orig, but there are 3 criteria where components are given a classification tag rather than being rejected. This will accept the same or more components than tedana_orig. This decision tree is used in a manuscript that is under review: Tahayori, B. et al "Robust-tedana: An automated denoising pipeline for multi-echo fMRI data" 2025

For each decision tree there is a json file which is what run as part of tedana, a png which is flow chart of the decision tree steps, and a tex file which was used to generate the flow chart.

People who use tedana software should cite: DuPre et al., (2021). TE-dependent analysis of multi-echo fMRI with tedana. Journal of Open Source Software, 6(66), 3669, https://doi.org/10.21105/joss.03669

In the tedana documentation, more information explaining the differences between these decision trees is in the FAQ: https://tedana.readthedocs.io/en/stable/faq.html
Code is at: https://github.com/ME-ICA/tedana/