Functional regions are delineated based on spatial interaction such as transportation, commuting and migration. A lot of delineation methods have been developed since 1960s. However, there is no agreement on the criteria to assess the validity of delineation result. In this study, we focus on the similarity between the delineation of functional regions and finding a community structure in a network. In the network sciences, a robustness approach has been proposed to find significant community structures. We employ this approach to the assessment of the validity of delineation results, and propose a method that assesses the robustness of functional regions against three types of perturbations: errors in observation of OD flow, variation of boundary of analysis area, and variation in parameters of delineation methods. If the original functional regions are robust to the perturbations, we conclude that the validity of delineated function regions is high. The results of case study show that our method can select one or a few valid delineation results from the masses of results. In addition, the results show that even when more than one results are selected, it is possible that there is hierarchical relations among selected results.