Domain adaptation on ABIDEII fmri data

Hello folks, just a new thread on the ABIDE dataset:

We are working on 3D CNN models transfer on structural and functional MRI data.
For the dataset, we have chosen ABIDE on CPAC as a benchmark. And after 6 months found nothing better than functional-connectivity classification without any domain adaptation (which is 0.70 on CV5).

In case you are working with ABIDE or domain adaptation on fmri - we will be happy to tell you about our failures and think of how to do something meaningful to the community on ASD detection.

BTW if someone has experience in working with raw ABIDE data from here and fmriprep launching on it, I will very appreciate if you will contact

1 Like

Hi Ekaterina! I am sure this was quite some time ago but I am looking into ABIDE I/II for ASD prediction using deep learning models. I am trying to run fmriprep but it seems that ABIDE isn’t really into BIDS format (I am currently getting error about “Repetition time is not set” but I am looking into ways to fix it).

Happy to hear more about success and failures that you had!