Appropriate Time Window for TENT Function

AFNI Message Board
1

Dear AFNI experts,

I am analyzing an fMRI experiment with 8 different 10-second blocks, each followed by a 6-second break, with a TR of 2 seconds. There’s no jittering, so the 16 seconds after each block onset coincide with the onset (0s) of the next block. I aim to calculate the average signal change over a specific interval (e.g., 6-10s after block onset) for each condition.

Following the program’s suggestion, I initially used the TENT function with parameters TENT(0,24,13) in 3dDconvolve, where the stop time c was set to the duration of the stimulus plus 14 seconds. However, I observed some collinearity issues, indicated by increased variance, when compared to using TENT(0,14,8), which ensures that the regressors do not overlap.

Given that I’m not interested in signal changes in the later period, is it more appropriate to use TENT(0,14,8)? Any suggestions or insights would be greatly appreciated.

Thanks in advance for your help!

2

Model construction is better guided by the underlying data-generating process, not solely by the investigator's interests. Before reducing the coverage period of the BOLD response, consider testing whether TENTzero(0,24,13) improves the model to some extent. Also, visually check the estimated hemodynamic responses and then decide the coverage period.

Gang Chen

3

Dear Gang,

Thank you for your helpful advice. I have tried using TENTzero(0,24,13), which outputs 11 time points, as it assumes the start and end time points are zero. Is this function technically equivalent to using TENT(2,22,11)?

Regarding the results, I found that the average signal change and variance were similar between TENT(0,24,13) and TENTzero(0,24,13). I also visually checked the estimated hemodynamic responses and observed that they tend to return to zero around 18-20 seconds after stimulus onset. Based on this, would it be reasonable to use TENTzero(0,18,10) to capture the response?

Thank you again for your insights and guidance.

4

TENTzero(0,24,13) is indeed equivalent to TENT(2,22,11). If BOLD responses generally recover fully within 18 - 20 seconds, using TENTzero(0,18,10) seems like a reasonable and appropriate choice.

Gang Chen