Prediction of vulnerable lines during cascading failures is an important issue for power networks. The identification of key lines may enable the application of targeted countermeasures and reduce cascading failures effects. Based on topological information and line power transmission capacity, we propose the use of cut-sets (CS) to quantity the line significance during cascading failures phenomenon. We calculate a CS-based measure using the Nagamochi-Ibaraki algorithm. Then, we compare the CS measure with other network-based and flow-based indices: the edge-betweenness centrality and the power flow centrality in a cascading failures framework. Simulation results show that the CS measure outperforms in predicting the edges significance for the cascading failures propagation path.