help
514 Subscribers 
help > Analysing resting-state fMRI data
Showing 1-2 of 2 posts
May 5, 2016 05:05 AM | Natasha Faria - VIT University
Analysing resting-state fMRI data
I am relatively new to working with fMRI data and have a basic
question regarding usage of connectivity toolboxes.
I have 2 groups of sample data: Patient and Control; both processed using the same steps; and to analyse the resultant data I have decided to compare the output from two different toolboxes.
I am comparing resultant analysis data from two toolboxes; one of which is CONN and the other being DPARSF.
I have used the 1st level bivariate correlation ROI-to-ROI analysis in CONN folllowed by 2nd level FDR corrected two sided p value computation to compare functional connectivity patterns between control and group.
In DPARSF, I have used only the Functional connectivity option to obtain the correlation coefficients for each subject. Following which I have used Matlab to compute Mean, Standard deviation for the two groups and subsequent two sided T test to compute the p value.
According to my knowledge, both the analysis methods appear to be similar, but the results from both are largely different; with DPARSF showing mainly only left ROI to right corresponding ROI higher functional connectivity.
Any clarification and assistance in the following matter would be greatly appreciated in order to understand this better.
I have 2 groups of sample data: Patient and Control; both processed using the same steps; and to analyse the resultant data I have decided to compare the output from two different toolboxes.
I am comparing resultant analysis data from two toolboxes; one of which is CONN and the other being DPARSF.
I have used the 1st level bivariate correlation ROI-to-ROI analysis in CONN folllowed by 2nd level FDR corrected two sided p value computation to compare functional connectivity patterns between control and group.
In DPARSF, I have used only the Functional connectivity option to obtain the correlation coefficients for each subject. Following which I have used Matlab to compute Mean, Standard deviation for the two groups and subsequent two sided T test to compute the p value.
According to my knowledge, both the analysis methods appear to be similar, but the results from both are largely different; with DPARSF showing mainly only left ROI to right corresponding ROI higher functional connectivity.
Any clarification and assistance in the following matter would be greatly appreciated in order to understand this better.
May 18, 2016 05:05 AM | Satoru Hiwa - Doshisha University
RE: Analysing resting-state fMRI data
Dear Natasha
I'm also facing a similar problem with you.
I'm also comparing CONN with DPARSFA regarding 1st-level ROI-to-ROI analysis, but their results are quite different.
I would think, this difference might be caused by difference in preprocessing procedures of them. CONN performs nuisance regression after slice-timing, realignment, (coregistration), normalization and smoothing, while DPARSFA does them in this order: slice-timing, realignment, (coregistration), nuisance regression, normalization and smoothing.
I have no idea about which is correct (or better), but it seems that the BOLD time course derived by DPARSFA is more noisy than that of CONN. I suppose that normalization and smoothing before nuisance regression may reduce noise in CONN.
How do you think about?
I would be grateful if someone who knew about two toolboxes could help me.
Best,
Satoru Hiwa
Originally posted by Natasha Faria:
I'm also facing a similar problem with you.
I'm also comparing CONN with DPARSFA regarding 1st-level ROI-to-ROI analysis, but their results are quite different.
I would think, this difference might be caused by difference in preprocessing procedures of them. CONN performs nuisance regression after slice-timing, realignment, (coregistration), normalization and smoothing, while DPARSFA does them in this order: slice-timing, realignment, (coregistration), nuisance regression, normalization and smoothing.
I have no idea about which is correct (or better), but it seems that the BOLD time course derived by DPARSFA is more noisy than that of CONN. I suppose that normalization and smoothing before nuisance regression may reduce noise in CONN.
How do you think about?
I would be grateful if someone who knew about two toolboxes could help me.
Best,
Satoru Hiwa
Originally posted by Natasha Faria:
I am relatively new to working with fMRI data
and have a basic question regarding usage of connectivity
toolboxes.
I have 2 groups of sample data: Patient and Control; both processed using the same steps; and to analyse the resultant data I have decided to compare the output from two different toolboxes.
I am comparing resultant analysis data from two toolboxes; one of which is CONN and the other being DPARSF.
I have used the 1st level bivariate correlation ROI-to-ROI analysis in CONN folllowed by 2nd level FDR corrected two sided p value computation to compare functional connectivity patterns between control and group.
In DPARSF, I have used only the Functional connectivity option to obtain the correlation coefficients for each subject. Following which I have used Matlab to compute Mean, Standard deviation for the two groups and subsequent two sided T test to compute the p value.
According to my knowledge, both the analysis methods appear to be similar, but the results from both are largely different; with DPARSF showing mainly only left ROI to right corresponding ROI higher functional connectivity.
Any clarification and assistance in the following matter would be greatly appreciated in order to understand this better.
I have 2 groups of sample data: Patient and Control; both processed using the same steps; and to analyse the resultant data I have decided to compare the output from two different toolboxes.
I am comparing resultant analysis data from two toolboxes; one of which is CONN and the other being DPARSF.
I have used the 1st level bivariate correlation ROI-to-ROI analysis in CONN folllowed by 2nd level FDR corrected two sided p value computation to compare functional connectivity patterns between control and group.
In DPARSF, I have used only the Functional connectivity option to obtain the correlation coefficients for each subject. Following which I have used Matlab to compute Mean, Standard deviation for the two groups and subsequent two sided T test to compute the p value.
According to my knowledge, both the analysis methods appear to be similar, but the results from both are largely different; with DPARSF showing mainly only left ROI to right corresponding ROI higher functional connectivity.
Any clarification and assistance in the following matter would be greatly appreciated in order to understand this better.