We are interested in the fully automatic delineation of the bladder in CT images in the frame of dose calculation for conformational radiotherapy. To this end we fit a series of 3D deformable templates to the contours of anatomical structures. The novelty of our approach resides in the ability to automatically adapt to different kinds of bladder images (homogenous, non–homogenous, contrasted or non–contrasted). The adaptation of the algorithm to inhomogeneities within the bladder improves the accuracy of the segmentation. We validate our approach on a database of tomodensitometric (CT) images of the lower abdomen of male patients.