There are numerous techniques for obtaining surface triangulations from 3D image data. Some applications require efficient triangulations, i.e. the source data should be represented as accurately as possible with a minimum number of triangles. This work is based on an existing algorithm, that adapts triangulation density to local surface curvature. To be useful in clinical practice, robustness has been improved and parameterization has been simplified by reducing the number of required arguments from twelve to a single detail control parameter. Triangulation times are 115 sec for 1.5mio source triangles and 3 sec for 60.000 triangles on a 3GHz P4.