This is empirically derived adjustment to account for the temporal smoothness. For simplicity it's factored into this variable, rather than dividing AdjEV().Ncols() by this factor.
On 16 Jun 2010, at 02:53, Jingqi Ao wrote:
> FSL expert:
> In MELODIC, the number of effective voxel for eigenspectrum adjustment seems to be:
> Number of effective voxel = Number of voxel / (2.5*Resolution Element)
> and Resolution Element = FWHMx * FWHMy * FWHMz.
> Why there is 2.5 in the denominator?
> Thank you.