Hello,

I am using the CONN toolbox and I am confused about the relationship between session-level (/data/ROI_Subject*_Session*.mat) and condition-level (results/preprocessing/ROI_Subject*_Condition*.mat) ROI timeseries outputs. I know that the first one represents the raw timeseries coming from the Setup step, while the second ones should be the timeseries after preprocessing and denoising. Nevertheless, there’s still something that is not so much clear to me.

My design consists of 4 sessions per subject, each one lasting 100 TR:

• Session 1 = DRUG_PRE


• Session 2 = PCB_PRE


• Session 3 = DRUG_POST


• Session 4 = PCB_POST

So I defined 4 conditions, each fully corresponding to one session (no overlap across sessions), in this way:

• DRUG_PRE → session 1 only (“condition spans the entire selected session” for Session 1 and “condition not present in the selected session” for Sessions2/3/4)


• PCB_PRE → session 2 only (same but for session 2)


• DRUG_POST → session 3 only (same but for session 3)


• PCB_POST → session 4 only (same but for session 4)

In this way, the final experimental design matrix is something like:
1 0 0 0

0 1 0 0

0 0 1 0

0 0 0 1

(where rows are conditions and columns are sessions, see the attached file).

At this point, I was interested in extracting the ROIs timeseries for each session/condition (since each condition is, in fact, the corresponding session), for some post-hoc analyses. At this point:

-            If I want to use the raw ROI_Subj..*_Session*.mat file, it’s sufficient to select for each session the data in position [X] corresponding to the timeseries of ROI [X], and I’ll obtain an array with size 100x1 (as expected) for each condition (or session, since each condition is a session)

-            If I want to use the denoised ROI_Subj..*_Condition*.mat file, I found out there are actually five files per Subject (Condition 000/001/002/003/004), which are all of them identical (same length, which is 400 – and not 100 TR -, and same content too).

So my questions are:

1. I know that this theme has been already discussed, but it’s still not clear the difference between condition 000 and all the others.
   
2. Why are all condition-level ROI outputs identical across conditions?
   
3. Why do ROI_Subject*_Condition*.mat appear to have length ~400 TR if each condition corresponds to a single 100-TR session? Since 400 is the sum of all the sessions, I supposed that CONN may internally concatenate denoised sessions, so I was thinking to extract denoised timeseries data by splitting from the Condition 000 file the 0-100, 101-200, 201-300 and 301-400 data to have single sessions timeseries. Is it correct? If not, what is the recommended way to extract ROI timeseries per experimental condition in a multi-session design like this?
   

I would really appreciate clarification on how CONN handles session vs condition-level ROI extraction.

Thank you!
Federica