Topic detection is an automatic method to extract topics in textual data, i.e., trending topic in social media. One of the recent topic detection methods is Eigenspace-based Fuzzy C-Means, which is a soft clustering-based topic detection method. In this method, the textual data are transformed into a lower-dimensional Eigenspace using truncated singular value decomposition. Fuzzy C-Means is performed on the Eigenspace to identify the memberships of each textual data to each cluster. Using these memberships, we extract the topics from textual data on the original space. In this paper, we use another approach to extract the topics by transforming back the centroids of the clusters into the positive subspace of the original space. Our simulations show that this new approach improves the old one regarding the topic interpretability in term of the coherence score. Moreover, this Eigenspace-based Fuzzy CMeans becomes better than both standard methods, i.e., nonnegative matrix factorization and latent Dirichlet allocation.