In order to properly define, determine and examine the topological structure of a dynamical system at the heart of the brain, there is a need to find the correct model and reconcile it with statistical data revealing latent effective relations between hidden states. Often the statistical data for that task is highly limited, and the reconciliation procedure sways with anomalies, when the chosen model allows such discrepancies. Recent works propose dynamical causal modeling in order to enforce causal and observational limitation for the inference process. Our work proposes two new addition to the literature; the more general framework of Dynamic Multiscale Networks (DMNs) coupled with the needed integration of structural information of a given brains anatomy, and the numerical analysis of the effectiveness of such a model. The unique properties of these models might naturally favor small worldness and thus they might be primary candidates for any "economic" networks such as the mammal brain. In particular, the general posing of DMNs allows for reconciliation of sparse functional relations and dense structural information.