help
help > RE: Help -- data results lost
Mar 10, 2022 12:03 AM | Alfonso Nieto-Castanon - Boston University
RE: Help -- data results lost
Hi Keith,
The results for the resting state runs should be identical (irrespective of whether additional task runs are included in your project or not). In particular any potential additional task effects included in the denoising step will not be present in the resting state runs, since denoising is run separately for each run and only regressors that are non-zero at least once within a run are considered. It's hard to tell but I would probably suggest to check whether perhaps your resting state data has been preprocessed twice (e.g. if after adding the new task runs preprocessing was run across the entire dataset), or perhaps whether the condition used to estimate connectivity during the rest runs may also be associated with at least some periods of your task runs?
Hope this helps
Alfonso
Originally posted by Keith Dodd:
The results for the resting state runs should be identical (irrespective of whether additional task runs are included in your project or not). In particular any potential additional task effects included in the denoising step will not be present in the resting state runs, since denoising is run separately for each run and only regressors that are non-zero at least once within a run are considered. It's hard to tell but I would probably suggest to check whether perhaps your resting state data has been preprocessed twice (e.g. if after adding the new task runs preprocessing was run across the entire dataset), or perhaps whether the condition used to estimate connectivity during the rest runs may also be associated with at least some periods of your task runs?
Hope this helps
Alfonso
Originally posted by Keith Dodd:
Hello, I am having an issue where I had solid
results, but now I cannot seem to recreate/find them again.
I had processed resting data in fmriprep. Uploaded that to Conn. Smoothed [8 8 8], denoised with normal denoising (including a parameter for each resting session). Then I had run analysis and found solid results!
So, I then also processed task data in fmriprep. Each run has 4 different tasks run up to 5 times per run. I imported that into Conn too. Reran denoising for all, and now not only have resting sessions as a variable, but also each task as a variable (e.g. 4 sessions, 4 tasks, 5 sections of each task in each run).
Now when I run my old same analysis for resting data, I no longer have the same significance. I lost several significant clusters, and those that still exist are smaller.
I do not feel that adding on analysis of task data should have deleted the significance I found in resting data, since those sessions/runs are independent of each other. Task results should not impact resting results?
Why did this occur? Is it because in denoising I am now not only controlling for 4 resting sessions, but also for 4 tasks over 4 sessions (16 more parameters) in denoising? How can I fix this so I can see my original resting results again? Maybe there is a way to denoise only resting sessions for resting data, and then denoise the task events for task data only?
Thanks!
I had processed resting data in fmriprep. Uploaded that to Conn. Smoothed [8 8 8], denoised with normal denoising (including a parameter for each resting session). Then I had run analysis and found solid results!
So, I then also processed task data in fmriprep. Each run has 4 different tasks run up to 5 times per run. I imported that into Conn too. Reran denoising for all, and now not only have resting sessions as a variable, but also each task as a variable (e.g. 4 sessions, 4 tasks, 5 sections of each task in each run).
Now when I run my old same analysis for resting data, I no longer have the same significance. I lost several significant clusters, and those that still exist are smaller.
I do not feel that adding on analysis of task data should have deleted the significance I found in resting data, since those sessions/runs are independent of each other. Task results should not impact resting results?
Why did this occur? Is it because in denoising I am now not only controlling for 4 resting sessions, but also for 4 tasks over 4 sessions (16 more parameters) in denoising? How can I fix this so I can see my original resting results again? Maybe there is a way to denoise only resting sessions for resting data, and then denoise the task events for task data only?
Thanks!
Threaded View
| Title | Author | Date |
|---|---|---|
| Keith Dodd | Mar 9, 2022 | |
| lidizpe | Jun 5, 2024 | |
| Keith Dodd | Mar 14, 2022 | |
| Alfonso Nieto-Castanon | Mar 14, 2022 | |
| Alfonso Nieto-Castanon | Mar 10, 2022 | |
| Keith Dodd | Mar 10, 2022 | |