cubic interpolation used to resample each voxel’s time-series data to a new rate such that the same time points are obtained for all voxels. This can be useful if the trial duration is not evenly divisible by the TR, which then handles any jitter between trial onsets and slice acquisition times.
The motivation for upsampling is to exploit the intrinsic jitter between the data acquisition and the experimental paradigm.
An equivalent functionality exists in matlab, see: https://github.com/kendrickkay/knkutils/blob/master/timeseries/tseriesinterp.m
The logic of this procedure was introduced in the following pre-print: https://www.biorxiv.org/content/10.1101/868455v1.full.pdf