Analysis of coexisting neuronal populations in optogenetic and conventional BOLD data

Modeling the hidden neuronal and hemodynamic sources of the blood-oxygenation-level-dependent (BOLD) signal allows for location-dependent parameter estimation that potentially contains more information about brain activation than the general linear model. Here we propose a generalization of the Buxton-Mandeville-Friston BOLD model for more than one neuronal population sharing common hemodynamics within a voxel and demonstrate that new hemodynamic response functions can result. Further, it is demonstrated that two neuronal contributions can be disentangled by parameter estimation from simulated data under certain conditions including differing neuronal efficacies and relaxation parameters. Finally, previously unexplained observations in optogenetic fMRI data are successfully modeled using this approach. This mapping of neuronal and hemodynamic sources could provide novel functional markers for not necessarily optogenetic clinical applications in a more specific way than standard clinical MRI protocols.