AdDressing Missing Nodes as Missing Data in Dynamic Causal Modeling

Dynamic Causal Modeling (DCM) uses dynamical systems to represent the high-level neural processing strategy for a given cognitive task. The logical network topology of the model is specified by a combination of prior knowledge and statistical analysis of the neuro-imaging signals. Parameters of this a-priori model are then estimated and competing models are compared to determine the most likely model given experimental data. Inter-subject analysis using DCM requires considerable judgement on the part of the experimenter to decide on the validity of assumptions used in the modeling and statistical analysis; in particular, the dropping of subjects with insufficient activity in a region of the model and ignoring activations not included in the model. This manual data filtering is required so that the model's network size is consistent across subjects. A solution to this problem would ease using DCM in population studies and reduce potential sources of experimental bias. The paper describes and compares three different approaches to allow inter-subject comparisons by treating variation in the network size as a missing data problem. These approaches are compared with respect to accuracy in classifying and predicting subject DCMs using simulated data.