help > Inconsistent results in 2018a vs. 2017f and when using converted Nifti files
Sep 12, 2018  04:09 PM | Kevin Nguyen
Inconsistent results in 2018a vs. 2017f and when using converted Nifti files
Hello all,

I've noticed that my resultant connectivity matrix is quite different depending on the version of CONN and also whether I'm loading DICOM files directly (allowing CONN to do the conversion with SPM12's spm_dicom_convert) or loading Nifti files that have been converted from DICOM using dcm2bids (which uses dcm2niix under the hood). 

The attached figure shows plots of a subject's connectivity matrix created with each combination of CONN version and DICOM conversion method. I used a custom atlas to define these ROIs. There's a marked difference between 2018a and 2017f, I'm guessing due to improvements made in the most recent version (the denoising seems to have improved?). There's also a difference between using dcm2niix and CONN's automatic conversion, but what's interesting is that the difference seems to be much greater in CONN 2017f compared to 2018a. 

Has something about how CONN reads Nifti files changed within the last update? Also, is there a changelog available that lists what kind of overall improvements have been made from the last version?

Thanks,
Kevin