Computation and data grids have encountered a large success among the scientific computing community in the past few years. The medical imaging community is increasingly aware of the potential benefit of these technologies in facing today medical image analysis challenges. In this paper, we report on a first experiment in deploying a medical application on a large scale grid testbed. Our pilot application is a hybrid metadata and image content-based query system that manipulates a large data set and for which image analysis computation can be easily parallelized on several grid nodes. We analyze the performances of this algorithm and the benefit brought by the grid. We further discuss possible improvements and future trends in porting medical applications to grid infrastructures.