Hi everyone,

I have a couple of questions that I would like to ask before starting my new project.

I have a longitudinal dataset of task-related fMRI acquistions. The subjects were acquired three times with a one-year gap between acquisition. During each acquisition subject performed a WM task. In particular they performed the experimental task with two different loads and a control task also with two different loads. Note that within each acquisitions, two task-related sessions were acquired. So to wrap it up I have four conditions (WM load 1, WM load 2, control load 1, control load 2) * 2 sessions * 3 different acquisitions.
We are interested in task-related connectivity, and one of the thing that we want to see is if the differential connectivity (with certain seeds) between experimental and control tasks changes between the different acquisitions.

The first question is about how to manage the different condition and the different acquisitions: in the setup tab, should I enter 6 different sessions, i.e. 2 sessions * 3 acquisitions ? Or should I concatenate the two sessions belonging to the same acquisition to have only 3 sessions corresponding to the 3 acquisitions ?

About the conditions: what seems most logic to me would be to create time-specific conditions, so that I would have WM load 1 -Time 1, WM load 2 - Time 1, etc; WM load 1 -Time 2, WM load 2 - Time 2, etc; WM load 1 - Time 3, WM load 2 - Time 3 etc. In this way, I'll be able to compare the same conditions in different time or test the interaction by conditions and time, isn't that the case ?

Finally, a more technical note: I have different onset times both across sessions and acquisitions of the same subject and across subjects. Moreover, we would like to only include correct trials. So, basically, each single onsets structure for each session and subject is unique. What would be the fastest way to include all these different onsets time in the conditions tab ?

Sorry for the long and maybe confusing mail and thank in advance for any suggestion

Alain

Hi Alain,

Regarding your first question about sessions/conditions, I would simply specify 6 sessions (2 sessions/runs * 3 acquisitions/days) in Setup.Basic, then enter the corresponding 6 functional timeseries (one per session) in Setup.functional, and last define your 12 conditions of interest (where each condition possibly spans multiple sessions) in Setup.conditions as the individual cells of your 2*2*3 within-subjects design (i.e. one condition for each combination of [WM/control], [load1/load2], and [time1/time2/ time3]; e.g. WM1_Time1, WM2_Time1, Control1_Time1, Control2_Time1, ... Control2_Time3)

Regarding your technical question (how to simply enter this condition info in CONN) you could create a single .csv or .txt file containing one row for each subject/session/condition combination specifying the corresponding onsets and durations associated with this condition in this subject and in this session, and then simply select 'import condition info from file' in Setup.conditions "condition tools" menu. See "help conn_importcondition" or this post http://www.nitrc.org/forum/message.php?m... for a more detailed description of the format of this file.

Alternatively, we are in the process of incorporating support in CONN for BIDS datasets, and that includes its own standardized way to specify onset/duration information for different subjects through a series of *_events.tsv files, so if you are already familiar with BIDS format and would like to give the new functionality a try please use the attached patch (this patch is for 17c, simply copy this file to the conn distribution folder overwriting the file with the same name there). When using the 'import condition info from file' option that will give you the choice to select between the CONN .csv/.txt format that I described above or BIDS-compatible .tsv files. 

Hope this helps
Alfonso

Originally posted by Alain Imaging:

Hi Alfonso, 

thanks a lot for your in depth answer.

I have just a couple of details to ask you about how to enter the covariate:

1) in the .csv or txt file, I guess that I should have two columns for each row: onsets and durations. So the very practical question is: should I enter the different onsets separated by commas, and separate the columns with a space ? Or is there any other specific format that I should use ?
2) about the order of the combinations conditions*subjects*sessions, is it correct that it should be: subj 1 - sess 1 - cond 1; subj 1 - sess 1 - cond 2; ...; subj 1 - sess 1 - cond n; subj 1 - sess 2 - con 1; ...; subj1 - sess k - cond n; subj 2 - sess 1 - cond 1; ...; subj s - sess k - cond n ?

Thanks again for your help!

Alain

Hi Alain,

The .csv/.txt file should have five columns separated by commas, describing the condition name, subject number, session number, list of onsets, durations, respectively. The onsets are specified as a list of numbers (separated by spaces) specified in seconds (english notation: decimals use dots). The file should also contain a header-line, and then an arbitrary number of rows (typically one for each subject/session/condition combination, but the order is arbitrary since the contents of each row are explicitly defined by the three columns).

For example, something like: (example contents of myconditions.csv file for two subjects, one session, two conditions)
   condition_name, subject_number, session_number, onsets, durations
   task, 1, 1, 0 50 100 150, 25
   task, 2, 1, 0 50 100 150, 25
   rest, 1, 1, 25 75 125 175, 25
   rest, 2, 1, 25 75 125 175, 25

Also note that you may leave the subject or session fields empty for any given row and that would be interpreted to indicate that the corresponding row applies to all subjects or sessions, respectively, so the above could be simplified as:
   condition_name, subject_number, session_number, onsets, durations
   task, , 1, 0 50 100 150, 25
   rest, , 1, 25 75 125 175, 25
 
Hope this helps
Alfonso

Originally posted by Alain Imaging:

Hi Alfonso, 
and thanks a lot for your answer ! I have just noticed that the link you provided in your previous answer had exactly the information I was looking for. Sorry for being obnoxious!

I have just realized one thing: being my three acquisitions one year apart, maybe it is not a very clever idea to use the same grey/white matter and csf mask for the three time points. Is there a way to enter session specific segmentation (I tried from the gui but the option "session specific ROI" cannot be selected for these ROIs. Is it advisable to use session specific segmentations ?

Alain