We introduce a methodology of removing the least important vortices in the surface representation of an anatomic model, and then filling the holes left behind with fewer polygons. Because the surfaces of organs and body structures can be highly convoluted, this decimation process needs not to be trapped by tortuous convolutions in the surface. Our theory is simple and intuitive. Fewer polygons are needed to accurately represent surface regions that are more flat than other regions. Conversely, more polygons are needed to represent surface regions characterized by high curvature. This theory leads us to compute a local curvature 'image' for the model and then search for those regions of low curvature as candidates for preferential decimation. For anatomic modeling of any given patient, it is necessary to only segment and model the organs of interest. Such patient-specific anatomic modeling has been demonstrated for use in virtual reality-based surgical planning systems by techniques which optimize the trade-off between model complexity and display/manipulation rates. We have described such a technique in this paper.