Health risks management such as epidemics study produces large quantity of spatio-temporal data. The development of new methods able to manage such specific characteristics becomes crucial. To tackle this problem, we define a theoretical framework for extracting spatio-temporal patterns (sequences representing evolution of locations and their neighborhoods over time). Classical frequency support doesn't consider the pattern neighbor neither its evolution over time. We thus propose a new interestingness measure taking into account both spatial and temporal aspects. An algorithm based on pattern-growth approach with efficient successive projections over the database is proposed. Experiments conducted on real datasets highlight the relevance of our method.