<p>Identifying key nodes in complex networks is crucial for understanding their robustness and vulnerability. Nevertheless, this task remains an open and challenging problem. Traditional centrality measures often rely on single-scale topological information, which limits their ability to fully characterize node importance. To address this issue, this paper proposes a multiscale structural contraction method based on discrete curvature for key node identification. From a geometric perspective, discrete curvature is employed to quantify the tightness of connections between nodes guiding the iterative contraction of local structures and progressively constructing a multiscale representation of the network. By integrating structural information preserved across different scales and mapping the aggregated multiscale evaluation back to the original graph, a comprehensive ranking of node importance is generated. Experimental results demonstrate that the proposed method outperforms existing baseline approaches on both real-world and synthetic networks, enabling more accurate and robust identification of structurally critical nodes.</p>