Résumé : The construction of articulatory models from medical images of the vocal tract, especially X-ray images, re lies on the application of an articulatory grid before deriving deformation modes via some factor analysis method. One difficulty faced with the classical semi-polar grid is that some tongue contours do not intersect the grid what gives rise to incomplete input vectors, and consequently poor tongue modeling in the front part of the mouth cavity which plays an important role in the articulation of many consonants. First, this paper describes preparation of data, i.e. drawing or tracking articulator contours, compensation of head movements and the construction of the adaptive polar grid. Then, the results of the principal component analysis are presented and compared with those obtained with the semi-polar grid.