SPM second-level con_0001.gii images are all zero?

Hello!

I am trying to analyse task-fMRI data in cifti format (HCP pipeline preprocessed). For reasons not mentioned here I want to do the analysis in SPM. I had assumed SPM could easily analyse surface .gii files. So I simply split the cifti into one volume and one surface file (converted to external binary, one image for each TR as required by SPM) and submit both separately to SPM. On the first level this creates beautiful outputs.

Yet, on the second-level the beta and con_000x.gii images are all zero!

Here is the thing: Given that the very same script can use volumetric input data easily and produce the expected results, I wonder what goes wrong when using .gii files as input. What is happening here?

  • Visual inspection of the .gii con images from the first level does not show anything unusual.

  • Further checks show that the input files exist, and they are not empty, or contain all zero or any NaN values.

  • I also don’t get an error or warning message from SPM.

I recently resurrected a thread from like seven years ago, but thought it may be better to create a new one (I followed the recommendations from the information in that thread and the email conversation on jisc regarding that topic).

Alternative question: Has anyone found a way to submit cifti files directly - without separating surface and volume - to SPM.