Many studies have characterized the effect of HRF variability across regions and subjects [45, 54], as well as the impact of HRF variability across layers [18, 46–50]. However, all of these studies investigated the impact of HRF variability in the context of detecting activation (and not in the context of characterizing functional connectivity). Furthermore, inter-subject and spatial variability of the HRF could potentially give rise to a scenario, wherein the BOLD fMRI time series from any given two regions are synchronized, while the underlying neural response is not, thus giving high correlation between BOLD signals, while the true correlation between latent neural variables may be low. The opposite scenario, wherein the underlying neuronal variables are synchronized, while the BOLD fMRI time series are not, is also equally possible (see Additional file 1: Fig. S1 for illustration of these scenarios). Therefore, we need to extract the underlying latent neural response to get reliable estimates of FC between layers of different regions. The readers are referred to Rangaprakash et al. for more details on the effects of HRF variability on functional connectivity [55].
