Hello,

I cannot advise on the first part of this, but afni_proc.py reports at the end of the processing how many time points were censored, along with fractions per stim class and overall.


Slice-based censoring would be possible using 3dREMLfit (assuming there were a reasonable way to detect slice-based motion used for censoring), but there is no such method available within the framework of afni_proc.py.

But slice-based censoring might come with issues.

Were slice-based censoring done in resting-state analysis, it is not clear how correlations would then be carried out, as censoring would vary across space. And the method would rely on reasonably good slice-based motion correction, so that resampling and smoothing operation would not contaminate neighboring slices with censored motion effects.

Plus, any linear regression model would almost have to be done on the original slice grid, as the locations, angles and spacing of slices would change on any move to standard space. And that would lead to an extra resampling of the data when subsequently going to standard space, adding blur and possibly requiring recomputation of blur estimation.

And while slice-based censoring could be applied in the linear regression model to compute betas, tracking degrees of freedom to report accurate t-stats and such might not be simple.

There seem to be a lot of complications with the notion of slice-based motion censoring at first thought.

But getting back to volumetric censoring, if SLOMOCO produces motion parameters over time, they could be passed along to afni_proc.py.

- rick