[Neuroimaging] downsampling without an anti-aliasing filter

[Blaise Frederick]

Hey Satra,

No, I don’t think you’re missing anything.  The only reason I can think that people would downsample fMRI data without aliasing is laziness from the bad old days of long TRs, where you might figure ‘well, cardiac and respiratory signals are aliased already, so why bother?”

Blaise


> On Sep 21, 2017, at 10:56 AM, Satrajit Ghosh <satra at mit.edu> wrote:
> 
> hi folks,
> 
> on the list and other places people often ask for scripts/tools for resampling or downsampling imaging data. my training in signal processing tells me that:
> 
> 1. upsampling and downsampling are different operations
> 2. if i downsample data without low-pass filtering at an appropriate cutoff i can introduce aliasing if there are frequencies present in the original data above the new nyqist rate.
> 
> for epi sequences where data are acquired in planes, i can easily understand doing this for slices. but i don't know why in imaging, many software downsample the in-plane samples without an anti-aliasing filter, especially as we are getting to higher sampling frequencies (smaller voxel sizes).
> 
> i would love to understand if i am missing something fundamental about imaging data or signal processing applied to imaging data.
> 
> cheers,
> 
> satra
> 
> _______________________________________________
> Neuroimaging mailing list
> Neuroimaging at python.org
> https://mail.python.org/mailman/listinfo/neuroimaging