3dClusterize and ClustSim.ACF


Is there a way to call the ClustSim.ACF*.1D file when using 3dClusterize? Specifically to input the appropriate -clust_nvox for the corresponding p-value, NN and 2sided?

I would like to bash 3dClusterize on a number of different subjects and would like to automatically use each corresponding -clust_nvox located in the /*.results/files_ClustSim folder. Or is it advised to just use the same -clust_nvox across all subjects? The values are not that far off from eachother.


This could be handled by scripting over each subject, where their input file is handled individually.

However, normally clusterizing with smoothing parameters is run on the group level, not the individual… For a given set of subjects on the same scanner/setup, then parameter values in the *1D file tend to be similar-- we just average across those, and apply that to the group level mask, say. An example of this is shown here:



I’m using fMRI in order to establish seeds for resting state. The plan is get the peak beta coefficient in 4 different regions (L/R M1 and PMd), after choosing an appropriate -pthr. Do you recommend another approach if clusterzing is typically done on the group level or is it okay to run it on individual subjects?

Would it be better to just use the peak t-stat?


So, you have several subjects processed and mapped to a common space, is that right? And are you aiming to place seeds and investigate maps, and then feedback and adjust seed locations? Or do you want to take an average of the time series in a particular region (e.g., L-M1) and then generate WB correlation maps and clusterize+combine those?

There is the GroupInCorr functionality that might help with this— you can investigate seedbased correlation across a group interactively in the GUI:
There is a demo of this AFNI Bootcamp, in “AFNI_InstaCorrDemo”, with instructions in the README*.txt, but we can discuss that more here, too.


Chronic stroke participants in the study underwent 5 weeks of rehabilitation and the primary objective is to assess pre to post changes in functional connectivity based on severity of impairment. In addition to resting state, participant performed a self-paced finger flexion/extension block paradigm.

The overall goal is to compare connectivity maps between ROI’s that were determined based on the fMRI analysis. As a result, all data (resting state and fMRI) is aligned to the participants mean T1 (pre and post). We processed the pre and post fMRI runs together (based on recommendations from the reviewers - originally we used the post fMRI maps) and the goal is to determined the peak activation voxel in the ipsilesional and contralesional M1 and PMd. Once we know the peak voxel, then we will created a 3x3 ROI around the voxel. We then created two whole brain connectivity map using the ipsilesional PMd and ipsilesional M1 as seeds (based on the peak voxel). We then will compare the FC between the generated ROI pairs (iPMd, iM1, cPMd, cM1) using the mean z-score in each ROI - all based on the fMRI activation map.

Because our hypothesis is that there will be differing changes in FC based on level of severity, we have a relatively small N in each subgroup. Therefore, we are performing the analysis on the single subject level rather than looking at group level maps.

With fMRI, I always have trouble choosing an appropriate -pthr, so I previously used Alphasim (an now 3dClustSim) to determine an adequate cluster size and -pthr on the single subject level. I also am never sure whether to use beta coefficients or t-stats, as I see them both used in the literature.

Based on this, what would you recommend we use for single subject -pthr? Is 3dClustSim not appropriate? Also, do you have an opinion on using the beta vs. t-stat?