In these last years, large quantity of spatio-temporal data stored leads to new needs such that management of natural risks, health or anthropogenic (e.g. understanding the dynamic of dengue epidemic). In this paper, we define a new theoretical framework for extracting spatio-sequential patterns. A spatio-sequential pattern is a sequence representing evolution of locations and their neighborhoods over time. We propose an efficient algorithm based on depth-first-search with successive projections over the database. We introduce a new interestingness measure taking into account both spatial and temporal aspects. Experiments are conducted on real datasets highlight the relevance of our method.