Hi there. I have never done biodistribution via pet/ct,  but I work quite a lot with different fluorescent images. In general I prefer to do whatever filters and adjustment my eye is accepting as improvment, getting some ROI coordinates from it (via segmentation or something like this) and then going back to original images to measure intensities and so one in given ROI from raw images. I am not sure if it can be applied in your case, but my knowledge of statistics and algorithms is telling me, that measruing processed signals can be very tricky and you must be 100% sure in what and how you change to do it. My general assumption is that only true professionals in this field, who write this algorithms (so you suspect them to be matematitians), can do it without fear to do mistakes.

By the way. I am more or less sure its not ok to use 3d Gaussian filter. To my knowledge (though I can be wrong) it messes with intensities. I am not sure what exactly is done in your case, but what I use (manually extraction of 2 differently blured by gaussian images) "sharpen" edges, but ruin intensity ration. Which is ok for segmentation, but not ok to measure.

So I wouldnt be surprised, that after this filter ration of low intensity/high intensity zones can change and I think its not ok to do it.

Though again - I am a biologist - can be wrong.

There is always a resolution/noise tradeoff when using filtering. If you use larger kernels, you will reduce more noise but small structures will loose contrast. If you are trying to quantify organ level distributions, then filtering is ok and won´t change your results. On the other hand, if you are quantifying small lesions, your measurements will get lower as you increase the filter size due to partial volume effects.