 |
 |
Dear Dr. Zeidman, Dear Dr. Razi and SPM list, I modified my questions that I post them last month. Hopefully, I will get a response this time. I have some technical, fundamental questions about spectral DCM (spDCM) for resting state fMRI. Please, accept my apologies for the long email. 1- How does spDCM estimate the sample cross spectra between ROIs? Based off what I understand that spDCM used Bayesian multivariate autoregressive models (fourth order), but MAR does not determine the hemodynamic parameters, is that correct? How does MAR specify the links between ROIs? How SPM12 applied Bayesian MAR to estimate the sample connectivity parameters (i.e., A matrix)? Is Harrison et al. (2003, Multivariate autoregressive modeling of fMRI time series) the right article for implementation MRA in SPM12? 2- Why in spDCM did you extract both sample cross spectra and cross-covariance functions as features of BOLD signals (Friston 2014)? Why did you not take the sample cross spectra or sample cross-covariance as BOLD features? Is that related to some properties of “pink noise” (your assumption about stationary noise)? And how does pink noise drive the changes in the resting state fMRI signal? Could you provide me some sentences in that direction, please? 3- What are the units of the amplitude and exponent parameters, if I understood correctly the values in DCM.Ep.a are in log scale? How can I get the exact values of DCM.Ep.a? Also, what are the units of hemodynamic parameters (transit, decay and epsilon)? Are these parameters also in log scale? I would like to plot DCM.Eps therefore I asking. 4- Are effective connectivity parameters of spDCM in Hz unit? However, Friston (2003) suggested that units of parameters are per second (Hz) in classical DCM. What about the units of spDCM parameters? Because in spDCM you are estimating the connectivity parameters from features of the BOLD signals, not from time series of BOLD signals. 5- How does spDCM predict cross-spectral density (CSD) of the neuronal fluctuations? If I am correct, spDCM used a fixed form over regions for neural fluctuation parameters. Why (in my case) the CSD of hidden neural states do not have similar values over regions (see figure_1)? 6- I would like to validate my model parameters. Can I do cross-validation on DCM parameters? If yes, how can I do that? 7- In my case, I have resting state fMRI data for 20 subjects. Each subject scanned four times (four conditions [two factors A and B, each factor two levels 1 and 2]). If you look at the figure_2, you will see that the hemodynamic parameters across subjects per condition. For example, Hypo region have positive values (cross four sessions) of transit parameter while other regions have negative values, what that mean? Also, the epsilon value is near to zero in condition A2. It will be helpful if someone could give me some hints about interpreting these hemodynamic parameters. Thank you very much for considering my questions, and look forward to some explanations. ######################################################################## To unsubscribe from the SPM list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=SPM&A=1