This paper proposes and analyzes a two-stage network topology processorthat considers a centralized or a distributed method to identify the HVDCgrids connectivity. In the first stage of the two-stage processor, the substationtopology is analyzed locally using an automated graph-based algorithm.Thereafter, the processed information is sent to the second stage todetermine the HVDC grid connectivity and detect islands. Two possible approachesnamely, centralized and distributed methods, are formulated andstudied for this second stage. In the centralized method, k-means clusteringis used to detect islands. In the distributed method, the informationprocessed locally at the substation is exchanged between neighboring substationsto realize the grid connectivity. For distributed islanding detection,the connectivity problem is formulated as a set of linear equations and solvediteratively using successive-over-relaxation method. The performance of twoproposed methods versus conventional one-stage method has been tested inan islanding scenario for a 5-terminal HVDC grid.