<p>Inferring both the dynamical parameters of nodes and the network structure from time series data is a central problem in complex network research. A key challenge is that node dynamics are influenced simultaneously by intrinsic properties and network interactions, making it difficult to disentangle these effects from observational data. To address this, we propose a two-stage inference framework that first estimates node dynamical parameters, which are then used to reconstruct the network structure. This approach allows for accurate separation of intrinsic dynamics and interaction effects and can be applied to both directed and undirected networks. Validation across multiple types of node dynamics and network topologies shows that our framework achieves superior topology reconstruction accuracy compared with existing methods, particularly when the network has a relatively uniform degree distribution and the time series contain nonlinear dynamics. The method is especially suitable for networks with known node dynamics and relatively homogeneous degree distributions.</p>