Dynamic Causal Modeling
Will Penny
UCLA
Wellcome Dept.of Imaging Neuroscience
The brain appears to adhere to two fundamental principles of functional organization: functional integration and functional specialization. Integration within and among specialized areas is mediated by effective connectivity. In this talk, we review two approaches to modeling effective connectivity from fMRI data: Structural Equation Models (SEMs) and Dynamic Causal Models (DCMs).
Both approaches employ model comparison frameworks that allow inferences about effective connectivity itself and about how connectivity may change under different perceptual or cognitive conditions. However, the two approaches are based on very different generative models.
In DCM, a distinction is made between the neuronal level and the hemodynamic level. Experimental inputs induce changes in effective connectivity at the level of neurodynamics, which in turn give rise to changes in the observed hemodynamic responses. In contrast, SEM assumes that changes in effective connectivity directly alter the covariance structure of the observed hemodynamic signals.
Because changes in effective connectivity in the brain occur at the neuronal level, DCM is generally considered the more appropriate framework for fMRI data. This review focuses on the underlying assumptions and limitations of each model and demonstrates their application to data from a study of attention to visual motion.
An example compares feedforward and reciprocal DCMs of activity in the dorsal visual system during attention to visual motion. In both models, photic stimulation enters V1, and motion and attention variables modulate the connection from V1 to V5. Bayesian model comparison favors the network with reciprocal connectivity.
