Hi Conn users,
We have a couple of questions concerning the graph theory analysis on the CONN toolbox.
The study aims to investigate the correlation between our variable of interest Y and the graph theory measures obtained from our task events. We have a group of 25 subjects and we defined a network including 20 ROIs.
The task includes 4 sessions, with three different conditions each (C1, C2 and, C3). Conditions were pseudorandomly presented.
After performing the preprocessing using fmriprep, we entered our functional data, anatomical scans and defined the GLM. Then we performed the denoising of the data.
- I have a question regarding the Bande Pass filter. We have fixed this filter to [0,008 – Inf]. Assuming that it is the best filtering for task event-related functional images. Could you provide me more information on how to choose this number? I am a bit lost.
When specifying the first level, I am not 100 % sure what type of analysis we should conduct to perform graph theory analyses. For now, we used the task modulation effect (gPPI) combined with the bivariate regression to look at the effective connectivity.
At the second level, we selected all the C1 conditions (related to the 4 different sessions) as well as the C2 conditions. We then created the following contrast [1 -1, 1 -1, 1 -1, 1 -1] to look at C2-C1. We also selected All subjects, our Y variable of interest and, set the contrast to [0 1] to investigate the correlation between Y and the differences in the connectivity between C2-C1.
Then to compute our matrix design for the graph theory, we used a z-score (threshold = 0,15 and keeping only the positive correlations) and visualized the results at p<0,05 FDR.
-Is that correct, or something seems weird?
When trying to export the Z_thr for each ROI pair and subject we ended with a 20*20*200 matrix. We assume that it is a ROIxROIxData point. With the data point representing the value for each subject and each session and each condition C1 and C2 (25*4*2).
-Is that correct?
-Is it possible to plot a heatmap representing the strength of the connectivity?
-Is it possible to plot the connexions? Or at least visualized them?
Thank you.
Best,
Remi J
We have a couple of questions concerning the graph theory analysis on the CONN toolbox.
The study aims to investigate the correlation between our variable of interest Y and the graph theory measures obtained from our task events. We have a group of 25 subjects and we defined a network including 20 ROIs.
The task includes 4 sessions, with three different conditions each (C1, C2 and, C3). Conditions were pseudorandomly presented.
After performing the preprocessing using fmriprep, we entered our functional data, anatomical scans and defined the GLM. Then we performed the denoising of the data.
- I have a question regarding the Bande Pass filter. We have fixed this filter to [0,008 – Inf]. Assuming that it is the best filtering for task event-related functional images. Could you provide me more information on how to choose this number? I am a bit lost.
When specifying the first level, I am not 100 % sure what type of analysis we should conduct to perform graph theory analyses. For now, we used the task modulation effect (gPPI) combined with the bivariate regression to look at the effective connectivity.
At the second level, we selected all the C1 conditions (related to the 4 different sessions) as well as the C2 conditions. We then created the following contrast [1 -1, 1 -1, 1 -1, 1 -1] to look at C2-C1. We also selected All subjects, our Y variable of interest and, set the contrast to [0 1] to investigate the correlation between Y and the differences in the connectivity between C2-C1.
Then to compute our matrix design for the graph theory, we used a z-score (threshold = 0,15 and keeping only the positive correlations) and visualized the results at p<0,05 FDR.
-Is that correct, or something seems weird?
When trying to export the Z_thr for each ROI pair and subject we ended with a 20*20*200 matrix. We assume that it is a ROIxROIxData point. With the data point representing the value for each subject and each session and each condition C1 and C2 (25*4*2).
-Is that correct?
-Is it possible to plot a heatmap representing the strength of the connectivity?
-Is it possible to plot the connexions? Or at least visualized them?
Thank you.
Best,
Remi J