help > Sliding window length
Showing 1-4 of 4 posts
Display:
Results per page:
Apr 30, 2024  09:04 AM | jasminstein
Sliding window length

Hi everyone,


I have tried setting up a sliding windows analysis with a window length of 120 seconds. Since my TR is two, this window should span 60 scans. However, examining the study design, CONN tells me that the created windows span 64 scans each. The same happens if I try to create one window per TR (conditions then span 4 scans each). Does anyone know why?


Thanks a lot!


Jasmin

May 6, 2024  04:05 PM | Alfonso Nieto-Castanon - Boston University
RE: Sliding window length

Hi Jasmin,


For continuous acquisition data all blocks defining the timing of individual conditions (including sliding windows) are convolved with the hemodynamic response function in order to better approximate the expected time and strength of BOLD responses associated with events occurring within each block. If you would like to skip this convolution step you may select in the Setup.Basic menu the optoin 'sparse acquisition' and that will disable all hrf-convolution steps in CONN.


Hope this helps


Alfonso


Originally posted by jasminstein:



Hi everyone,


I have tried setting up a sliding windows analysis with a window length of 120 seconds. Since my TR is two, this window should span 60 scans. However, examining the study design, CONN tells me that the created windows span 64 scans each. The same happens if I try to create one window per TR (conditions then span 4 scans each). Does anyone know why?


Thanks a lot!


Jasmin



 

May 10, 2024  08:05 AM | jasminstein
RE: Sliding window length

Thanks, Alfonso, that helps! I am dealing with resting-state data here, not with specific events. Would this mean that setting the acquisition to sparse is more appropriate here?

May 11, 2024  04:05 PM | Alfonso Nieto-Castanon - Boston University
RE: Sliding window length

Hi Jasmin


For resting-state data the default 'continuous' acquisition should be more appropriate (and the corresponding smoothing and delay of the sliding windows caused by the hrf convolution step is also perfectly fine in order to better capture the BOLD response associated with each specific temporal window)


Best


Alfonso


Originally posted by jasminstein:



Thanks, Alfonso, that helps! I am dealing with resting-state data here, not with specific events. Would this mean that setting the acquisition to sparse is more appropriate here?