Dear Mr. Winkler,
Thank you very much for your answer. It was very informative. But unfortunately I could not get the answer to my second question:
2- Would you please let me know in case we use dual regression for feature extraction how are we supposed to extract that same features for test data? should we just use the group ICA template we had created using train data and just change the directories in list.txt to address test data?
To make myself clear, suppose I have some feature extracted from the ICs calculated using Group_ICA in melodic and regressed to correspond to subjects using dual regression.
Now if I want to have the same features extracted for a new data set ( which is my test data) how am I supposed to calculate corresponding ICs? I believe that I shouldn't run single subject ICA for each subject in test data as the ICs calculated will differ from the original group_ica ICs used in training data and features won't be comparable. I read in articles that they use the group_ICA template of the traing data and use dual regression to form subject level ICs. but to do so we need some data under the test data directory like the one used in training data: ..../x/x_ICA.ica/reg_standard/filtered_func_data.nii.gz
and we don't have this file for test data.
Would you please help me figure this out? any help will be highly appreciated.
Sincerely,
Shiva
|