 |
 |
Dear Peter, Thank you very much for the guidance in DCM analyses. We found between-group differences using gPPI method, and are interested in further explaining the task-modulation on amygdala-PFC connection -- especially the amygdala-->PFC and amygdala<-- PFC connections. However, we got some DCM findings that are inconsistent with gPPI analyses results: for gPPI: (1) patients versus controls showed decreased gPPI to amygdala-PFC connection; (2) a negative correlation between gPPI to amygdala-PFC connection and behavioral performance was found in controls only. for DCM: (1) no significant between-group difference of DCM parameters was found; (2) some positive/negative correlations between DCM parameters (intrinsic connections, and task-modulated amygdala-->amygdala connection, NOT amygdala-->PFC or amygdala<--PFC) and behavioral performance were found in patients only. We are now confused on how to interpret the findings. Could you please give us some hints about why there are such big differences between gPPI and DCM findings? Thank you very much! Bests, Delin On Thu, Jan 12, 2017 at 9:11 AM, 孙得琳 <[log in to unmask]> wrote: > Dear Peter, > > Thank you very much for the clarifications! > > Bests, > Delin > > > On Thu, Jan 12, 2017 at 6:27 AM, Zeidman, Peter <[log in to unmask]> > wrote: > >> Dear Delin Sun >> >> >> >> Thank you for the answers. Should I also include the 6 head motion >> parameters in GLM? It seems that SPM12 manual's example did not include >> head motion regressors in GLM. >> >> >> >> Yes, it’s always good to model known contributions to the signal. Motion >> will then be regressed out during ROI extraction (thanks to your Effects of >> Interest F-contrast telling SPM which columns are interesting). >> >> >> >> Thank you for the suggestion. I found in the SPM mailing list that Donald >> McLaren suggested to use his gPPI functions for concatenation. Is there any >> difference between your method and Donald's way? >> >> >> >> The SPM function I mentioned (spm_fmri_concatenate) adjusts the high pass >> filter and temporal autocorrelation model to account for the session >> concatenation. I haven’t seen Donald’s script, but I would be surprised if >> it does this. >> >> >> >> I am sorry for the confusion. I found in the SPM12 manual's example that >> the time series was extracted from VOI defined by "i1 AND i2" where i1 is >> the cluster of SPM outputs and i2 is a sphere covering the predefined ROI. >> What I want to do is to just extract time series from i2. Is it a correct >> way? >> >> >> >> Yes that’s fine – if you’re using the “eigenvariate” button in the GUI, >> first set the p-value to p=1 (to include all voxels), then click >> eigenvariate and select sphere. If you’re using the batch, just select >> ‘New: Sphere’ from the ‘Region(s) of Interest’ option and set the >> coordinates. The expression would then just be ‘i1’. Let me know if you >> need more clarification on these options. >> >> >> >> Thank you for the clarification. About C-matrix, if I understand it >> correctly, only column 1 is related with Task driving input, right? >> >> >> >> That’s right – it’s one column per input, one row per region. So the >> first column will be Task and the second column will be Emotion. >> >> >> >> All the best, >> >> Peter >> > >