Dear Irina,
I do not have a good explanation for your findings. I have however seen cases where data looked perfectly normal in terms of intensities but parts of each image were noisier than others. Visual inspection, tsdiffana et al of these raw data didn't reveal anything suspicious.
When looking at the data, you should set interpolation to "nearest neighbour" and orientation to "voxel space". In addition to the beta images, you might want to look at the RPV and ResMS images. These tell you about the smoothness of the error estimates and the model fit. Also it might be helpful to compute a standard deviation image of your time series. You can easily do that using ImCalc: select raw or smoothed data, enter "std(X)" as formula, set interpolation to nearest neighbour and "Read data into data matrix" to "Yes".
Instead of using "Display" you might want to use "Check Reg" and look at more than one image at the same time.
Best,
Volkmar
|