Detailed Description
After fitting the regression model with the PetCO2 signal plus possible additional (nuisance) regressors, we might want to do extra steps with the signal fitted with the entire model, or the signal only corresponding to the PetCO2 changes, or the 'denoised' signal that is the signal corresponding to the PetCO2 changes plus all the signal sources that are not modelled (i.e. unexplained variance or residuals).
Context / Motivation
We could use these signals as inputs to additional analyses, for instance clustering, parcellation, etc. In particular, Moia et al., 2021 showed that the denoised signal is less affected by motion artefacts (DVARS shows less correlation with FD).
This option will give users a cleaner CVR-related signal to perform extra analyses.
Possible Implementation
This implementation would involve fitting the model back with the estimated coefficients for each part of the regression model, and subtract each component from the input data in order to generate the different signals. This is similar to 3dSynthesize in AFNI.
Detailed Description
After fitting the regression model with the PetCO2 signal plus possible additional (nuisance) regressors, we might want to do extra steps with the signal fitted with the entire model, or the signal only corresponding to the PetCO2 changes, or the 'denoised' signal that is the signal corresponding to the PetCO2 changes plus all the signal sources that are not modelled (i.e. unexplained variance or residuals).
Context / Motivation
We could use these signals as inputs to additional analyses, for instance clustering, parcellation, etc. In particular, Moia et al., 2021 showed that the denoised signal is less affected by motion artefacts (DVARS shows less correlation with FD).
This option will give users a cleaner CVR-related signal to perform extra analyses.
Possible Implementation
This implementation would involve fitting the model back with the estimated coefficients for each part of the regression model, and subtract each component from the input data in order to generate the different signals. This is similar to 3dSynthesize in AFNI.