Dynamic causal modelling for fMRI: A two-state model

Publikation: Beitrag in FachzeitschriftForschungsartikelBeigetragenBegutachtung


  • A. C. Marreiros - , University College London (Autor:in)
  • S. J. Kiebel - , University College London (Autor:in)
  • K. J. Friston - , University College London (Autor:in)


Dynamical causal modelling (DCM) for functional magnetic resonance imaging (fMRI) is a technique to infer directed connectivity among brain regions. These models distinguish between a neuronal level, which models neuronal interactions among regions, and an observation level, which models the hemodynamic responses each region. The original DCM formulation considered only one neuronal state per region. In this work, we adopt a more plausible and less constrained neuronal model, using two neuronal states (populations) per region. Critically, this gives us an explicit model of intrinsic (between-population) connectivity within a region. In addition, by using positivity constraints, the model conforms to the organization of real cortical hierarchies, whose extrinsic connections are excitatory (glutamatergic). By incorporating two populations within each region we can model selective changes in both extrinsic and intrinsic connectivity. Using synthetic data, we show that the two-state model is internal consistent and identifiable. We then apply the model to real data, explicitly modelling intrinsic connections. Using model comparison, we found that the two-state model is better than the single-state model. Furthermore, using the two-state model we find that it is possible to disambiguate between subtle changes in coupling; we were able to show that attentional gain, in the context of visual motion processing, is accounted for sufficiently by an increased sensitivity of excitatory populations of neurons in V5, to forward afferents from earlier visual areas.


Seiten (von - bis)269-278
PublikationsstatusVeröffentlicht - 1 Jan. 2008
Extern publiziertJa

Externe IDs

PubMed 17936017



  • Dynamic causal modelling, Functional magnetic resonance imaging, Neural mass model