Dear AFNI expert (Daniel Glen):

I am trying to figure out some non-subjective way to censor TRs from messy data (i.e., not visual inspection). We currently just censor TRs whose signal is greater than some stdev (e.g. 5) of the mean, but the appropriate value of this varies from subject to subject, and right now selecting the appropriate value just becomes another subjective evaluation of the data.

I have been thinking of playing around with 3dBrickStat to calculate percentiles to see if this can give me some objective threshold by which to determine the censor value for a given dataset.

Because I would be calculating percentiles on 4D data, i would need to be running the program on multiple sub-bricks, which i know is not a current function.

Am wondering if there's some other way to do something like this in AFNI, or if anyone has developed a non-visual inspection based method for data censoring?

Thanks so much,

Dharol