We are doing a rt-fMRI-neurofeedback experiment on RELAB. We have a block design where participants are shown exciting food pictures. In one condition they have to admit/increase their desire/appetite and in the other condition they have to downregulate it. At the end of every trial they see a feedback-thermometer. We are also recording GSR data as a measure of emotional arousal with a setup from ADInstruments, which we want to analyze with the fMRI data (SN/VTA). This should be done in a similar manner as in Sulzer et al. (2013) (1). They fitted the GSR data to a GLM based on task onset and extracted the beta values of the SCR data from each run. The beta values were then put into an ANCOVA with the fMRI betas to determine wether the covariation between SCR and fMRI beta (SN/VTA) varied between groups. So our goal is to extract SCR beta values also for each run and put it into an ANCOVA with fMRI betas to determine wether the covariaton varies between the admit and neutralize condition.
The protocol is as follows: after baseline-fixation cross they get the admit-instruction followed by four 5sec-stimuli (food pictures) which come directly after each other. This is followed by 10sec fixation cross. Then the instruction "neutralize" appears, followed again by four 5sec-stimuli where they have to downregulate. Such a trial is repeated 5 times in no-feedback runs and 7 times in feedback runs (I have an image available for a graphical illustration of our design, but I can't find out how to upload to this forum entry, if I press "insert image" the page asks me to give a web link. ).
We want to extract one beta value for every condition (admit vs neutralize) for each run. So for the moment we just want to apply a first-level model per run. The markerfile is a .log-file which is imported in matlab with a script. At the moment we have coded an onset per 4 pictures, so the stimulus is defined as a 20sec epoch with 4 pictures.
So our question is which model would you recommend for our design? For now we were dealing with a GLM because Sulzer et al. (2013) also used a GLM and after reading the manual we thought it was the right way to go. But after going through the forum we are quite unsure wether a GLM fits to our design or if we should better use the DCM approach or due to the long duration of our stimuli (20sec) even use the SF model? Reading the forum we found that GLM would maybe not be best for longer events and that you would recommend and SF model for block design. There are also some recommendations for the DCM model for longer stimulus presentations or the SF-DCM approach.
So we are looking forward to hearing from you and thank you in advance!
Best regards,
Roland
(1) Sulzer, J., Sitaram, R., Blefari, M.L., Kollias, S., Birbaumer, N., Stephan, K.E., Luft, A., Gassert, R. (2013). Neurofeedback-mediated self-regulation of the dopaminergic midbrain. NeuroImage, 83, 817-825.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
GLM makes the assumption that the elicited SCR is time-locked to "something" that is known, e.g. an external event, or a subject's button press. It will miss (or mis-assign) SCRs that are not time-locked.
DCM is a more flexible way of dealing with unknown timings of responses. If you expect more than 1-2 SCR per event (have a quick look at your raw data), the parameter space gets difficult to deal with, and then the SF model is more constrained (it is actually the same model but the dispersion of the responses is fixed, thus reducing parameter space).
Hope this helps
Dominik
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi Prof. Bach & Mr. Moser
We are doing a rt-fMRI-neurofeedback experiment on RELAB. We have a block design where participants are shown exciting food pictures. In one condition they have to admit/increase their desire/appetite and in the other condition they have to downregulate it. At the end of every trial they see a feedback-thermometer. We are also recording GSR data as a measure of emotional arousal with a setup from ADInstruments, which we want to analyze with the fMRI data (SN/VTA). This should be done in a similar manner as in Sulzer et al. (2013) (1). They fitted the GSR data to a GLM based on task onset and extracted the beta values of the SCR data from each run. The beta values were then put into an ANCOVA with the fMRI betas to determine wether the covariation between SCR and fMRI beta (SN/VTA) varied between groups. So our goal is to extract SCR beta values also for each run and put it into an ANCOVA with fMRI betas to determine wether the covariaton varies between the admit and neutralize condition.
The protocol is as follows: after baseline-fixation cross they get the admit-instruction followed by four 5sec-stimuli (food pictures) which come directly after each other. This is followed by 10sec fixation cross. Then the instruction "neutralize" appears, followed again by four 5sec-stimuli where they have to downregulate. Such a trial is repeated 5 times in no-feedback runs and 7 times in feedback runs (I have an image available for a graphical illustration of our design, but I can't find out how to upload to this forum entry, if I press "insert image" the page asks me to give a web link. ).
We want to extract one beta value for every condition (admit vs neutralize) for each run. So for the moment we just want to apply a first-level model per run. The markerfile is a .log-file which is imported in matlab with a script. At the moment we have coded an onset per 4 pictures, so the stimulus is defined as a 20sec epoch with 4 pictures.
So our question is which model would you recommend for our design? For now we were dealing with a GLM because Sulzer et al. (2013) also used a GLM and after reading the manual we thought it was the right way to go. But after going through the forum we are quite unsure wether a GLM fits to our design or if we should better use the DCM approach or due to the long duration of our stimuli (20sec) even use the SF model? Reading the forum we found that GLM would maybe not be best for longer events and that you would recommend and SF model for block design. There are also some recommendations for the DCM model for longer stimulus presentations or the SF-DCM approach.
So we are looking forward to hearing from you and thank you in advance!
Best regards,
Roland
(1) Sulzer, J., Sitaram, R., Blefari, M.L., Kollias, S., Birbaumer, N., Stephan, K.E., Luft, A., Gassert, R. (2013). Neurofeedback-mediated self-regulation of the dopaminergic midbrain. NeuroImage, 83, 817-825.
Found out how to post the image, so please find attached a graphical illustration of our protocol. :)
Best regards,
Roland
Hi Roland
GLM makes the assumption that the elicited SCR is time-locked to "something" that is known, e.g. an external event, or a subject's button press. It will miss (or mis-assign) SCRs that are not time-locked.
DCM is a more flexible way of dealing with unknown timings of responses. If you expect more than 1-2 SCR per event (have a quick look at your raw data), the parameter space gets difficult to deal with, and then the SF model is more constrained (it is actually the same model but the dispersion of the responses is fixed, thus reducing parameter space).
Hope this helps
Dominik
Dear Dominik
Thank you very much for your answer and explanation! I think it helps a bit. :)
Have a nice weekend!
Best regards,
Roland