help
help > RE: two sample t test result (seed-to-voxel analysis)
Jun 9, 2020 12:06 AM | Alfonso Nieto-Castanon - Boston University
RE: two sample t test result (seed-to-voxel analysis)
Dear Xinyuan,
Everything looks fine from your description and results. Unlike in ROI-to-ROI analyses, where connectivity between a seed and itself is not really meaningful (it is always exactly r=1), in seed-to-voxel analyses the connectivity between a seed ROI and voxels within the same target area is perfectly meaningful and informative, and it is typically interpreted in terms of the degree of regional similarity or homogeneity of this seed/ROI (higher connectivity means that different voxels within the ROI show more similar BOLD timeseries).
Hope this helps
Alfonso
Originally posted by Dong An:
Everything looks fine from your description and results. Unlike in ROI-to-ROI analyses, where connectivity between a seed and itself is not really meaningful (it is always exactly r=1), in seed-to-voxel analyses the connectivity between a seed ROI and voxels within the same target area is perfectly meaningful and informative, and it is typically interpreted in terms of the degree of regional similarity or homogeneity of this seed/ROI (higher connectivity means that different voxels within the ROI show more similar BOLD timeseries).
Hope this helps
Alfonso
Originally posted by Dong An:
Dear Alfonso and CONN experts,
First, I very appreciate your great contribution to CONN toolbox. It is great!
I am new at using CONN, and I am confused about my two sample t test result. I selected right amygdala as ROI, and used ROI-to-voxel analysis type to compare patient and control resting state functional connectivity. I think if amygdala is a major cluster in both patient and control groups, it should not appear in 2 sample t test result.
But in my result, it is surprising that amygdala is involved in the biggest clusters and the peak location is at right amygdala.
Here is how I did:
I imported fmriprep preprocessed data (52 subjects: 31 patients 21 controls) into CONN, for right amygdala ROI, I used the Harvard-Oxford structural atlas by default in CONN. For 1st level covariates, I removed all covariates (because the fmriprep preprocessed data have been denoised using ICA-aroma). For 2nd level covariates, I added "patient group"(enter 1 for 31 patients and enter 0 for 21 controls), and added "control group(enter 0 for 31 patients and enter 1 for 21 controls). In denoising section, I added only CSF (5p) and white matter (5p) as covariates to regress, I did not add effect of rest. Then I click linear detrending, pass filtering: [0.01 inf] after regression. Then I did 1st level analysis (the setting is: functional connectivity weighted GLM correlation(bivariate) no weighting ), and 2nd level analysis design is: patient, control [1 -1] rest [1] seeds/sources: right amygdala [1].
Please see attached screenshot of "result explorer"; after t sample t test, the peak coordinate of the biggest cluster is (22,-14,-12), and please see attached .txt file containing all subregions of this cluster. I do not understand why right and left amygdala is still involved in this biggest cluster after 2 sample t test. Also, when I tried adding age or sex as covariates: patient, control, age [1 -1 0], amygdala still appear. Could you please tell me where I was wrong?
Thank you in advance!
Xinyuan
First, I very appreciate your great contribution to CONN toolbox. It is great!
I am new at using CONN, and I am confused about my two sample t test result. I selected right amygdala as ROI, and used ROI-to-voxel analysis type to compare patient and control resting state functional connectivity. I think if amygdala is a major cluster in both patient and control groups, it should not appear in 2 sample t test result.
But in my result, it is surprising that amygdala is involved in the biggest clusters and the peak location is at right amygdala.
Here is how I did:
I imported fmriprep preprocessed data (52 subjects: 31 patients 21 controls) into CONN, for right amygdala ROI, I used the Harvard-Oxford structural atlas by default in CONN. For 1st level covariates, I removed all covariates (because the fmriprep preprocessed data have been denoised using ICA-aroma). For 2nd level covariates, I added "patient group"(enter 1 for 31 patients and enter 0 for 21 controls), and added "control group(enter 0 for 31 patients and enter 1 for 21 controls). In denoising section, I added only CSF (5p) and white matter (5p) as covariates to regress, I did not add effect of rest. Then I click linear detrending, pass filtering: [0.01 inf] after regression. Then I did 1st level analysis (the setting is: functional connectivity weighted GLM correlation(bivariate) no weighting ), and 2nd level analysis design is: patient, control [1 -1] rest [1] seeds/sources: right amygdala [1].
Please see attached screenshot of "result explorer"; after t sample t test, the peak coordinate of the biggest cluster is (22,-14,-12), and please see attached .txt file containing all subregions of this cluster. I do not understand why right and left amygdala is still involved in this biggest cluster after 2 sample t test. Also, when I tried adding age or sex as covariates: patient, control, age [1 -1 0], amygdala still appear. Could you please tell me where I was wrong?
Thank you in advance!
Xinyuan
Threaded View
| Title | Author | Date |
|---|---|---|
| Dong An | Jun 3, 2020 | |
| Alfonso Nieto-Castanon | Jun 9, 2020 | |
| Dong An | Jun 10, 2020 | |