minimum_image_regression(optcom_ts, mmix, mask, comptable, io_generator)[source]

Perform minimum image regression (MIR) to remove T1-like effects from BOLD-like components.

While this method has not yet been described in detail in any publications, we recommend that users cite 1.

  • optcom_ts ((S x T) array_like) – Optimally combined time series data

  • mmix ((T x C) array_like) – Mixing matrix for converting input data to component space, where C is components and T is the same as in optcom_ts

  • mask ((S,) array_like) – Boolean mask array

  • comptable ((C x X) pandas.DataFrame) – Component metric table. One row for each component, with a column for each metric. The index should be the component number.

  • io_generator (tedana.io.OutputGenerator) – The output generating object for this workflow


Minimum image regression operates by constructing a amplitude-normalized form of the multi-echo high Kappa (MEHK) time series from BOLD-like ICA components, and then taking voxel-wise minimum over time. This “minimum map” serves as a voxel-wise estimate of the T1-like effect in the time series. From this minimum map, a T1-like global signal (i.e., a 1D time series) is estimated. The component time series in the mixing matrix are then corrected for the T1-like effect by regressing out the global signal time series from each. Finally, the multi-echo denoising (MEDN) and MEHK time series are reconstructed from the corrected mixing matrix and are written out to new files.

This function writes out several files:




T1-like effect


T1-corrected BOLD (high-Kappa) time series


Denoised version of T1-corrected time series


T1 global signal-corrected components


T1 global signal-corrected mixing matrix



Kundu, P., Brenowitz, N. D., Voon, V., Worbe, Y., Vértes, P. E., Inati, S. J., … & Bullmore, E. T. (2013). Integrated strategy for improving functional connectivity mapping using multiecho fMRI. Proceedings of the National Academy of Sciences, 110(40), 16187-16192.