<p>We consider the inverse problem of identifying a sound-soft inclusion in a bounded two-dimensional domain from interior measurements of transient acoustic waves over a finite time interval. By exploiting interior wavefield data-rather than boundary observations-we achieve enhanced identifiability, provided that the observation time exceeds the maximal distance between the sensing region and any candidate inclusion boundary. We formulate the reconstruction as a least-squares misfit between simulated and observed interior fields and derive its leading-order asymptotic expansion under the insertion of a vanishing Dirichlet inclusion. This leads to an explicit expression of the topological derivative in two dimensions, which is then used to design a “one-shot” imaging algorithm. In this approach, the topological gradient is computed from a single background forward solve and one adjoint solve per source, and its most negative peaks accurately indicate potential inclusion centers. We establish uniform-in-time error bounds for the asymptotic expansion, exploit sharp space-time regularity to justify pointwise evaluations, and address practical issues such as measurement noise and multiple inclusions. Numerical experiments involving various inclusion shapes, contrasts, and noise levels demonstrate the rapid, robust, and accurate localization capabilities of the proposed method. Overall, this approach provides a theoretically rigorous and computationally efficient framework for fast obstacle detection using interior wavefield data.</p>