Condition specific HRF

[Edouard2laire]

Hello,

First, thanks a lot for creating this tool. I am interested in the brain response to stimuli during sleep.
So I have my task paradigm (random stimulation during the sleep) and the brain sleep stage (wake, N1, N2, N3).

My question is: Does the HRF to the same stimulus change with sleep stage?

GMLsingle allows for a different HRF per voxel, but is there a way to have a different HRF per condition?

I was thinking of two solutions:

1. Fit one GLM per sleep stage, specifying only the stimuli happening during that specific sleep stage. But then the rest period would not really be rest.
2. Split the night into runs, splitting the night so each run corresponds to one sleep stage. Then HRFindexrun would indicate one HRF per run/sleep stage, answering the question.

Is there any risk in doing that?

Thanks a lot
Edouard

[kendrickkay]

Sorry for the delay.

"GMLsingle allows for a different HRF per voxel, but is there a way to have a different HRF per condition?"
==>
No, not in any straightforward way.

"Split the night into runs, splitting the night so each run corresponds to one sleep stage. Then HRFindexrun would indicate one HRF per run/sleep stage, answering the question."
==>
I think that's the most straightforward solution. Basically, if I understand correctly, you could simply divide your overall experiment into "sub experiments" (one for each sleep stage). And then analyze each sub-experiment separately (giving GLMsingle the data for one sub-experiment at a time).

[Edouard2laire]

Thanks a lot for your answer!!

Just to be sure, I understand correctly. Let's say I have only 2 sleep stage occuring like this: A B A B A.

1/ Using the entire data ( A B A B A) as one run doesn't work as the same HRF is used

2/ I can run with 3 run A data = {A1, A2, A3} and then with the 2 B run data = {B1, B2}. Then I have one hrf for A and one hrf for B

3/ i can run with 5 run as input {A1, B1, A2, B2, A3}. Then HRFindexrun should give me one HRF per run; so here 5 HRF.

Am I correct? From your message, is solution 2 better?

Edouard

[kendrickkay]

Just to be sure, I understand correctly. Let's say I have only 2 sleep stage occuring like this: A B A B A. 1/ Using the entire data ( A B A B A) as one run doesn't work as the same HRF is used 2/ I can run with 3 run A data = {A1, A2, A3} and then with the 2 B run data = {B1, B2}. Then I have one hrf for A and one hrf for B 3/ i can run with 5 run as input {A1, B1, A2, B2, A3}. Then HRFindexrun should give me one HRF per run; so here 5 HRF.

This is all correct. Regarding "HRFindexrun", it is just provided for informational purposes (and since it is easy to compute). GLMsingle doesn't actually use HRFs per run; rather it uses a single HRF across all runs. So, I would recommend that #2 is the best approach here.

[Edouard2laire]

great. Thanks a lot !