Collision detection is the key technology in nose augmentation surgery simulation system, which can avoid incorrect intersection between bones and the implant model. In this paper, we present a sphere-tree-based hierarchical collision detection algorithm (STHCD) for surgery simulation. Each model is represented as a clustered hierarchy of sphere tree (CHST). Graphics processing unit (GPU)-based occlusion queries were used for fast collision culling between massive models. Furthermore, we are able to generate these hierarchies and perform collision queries using out-of-core techniques on all triangulated models. Experimental results show that STHCD performs real-time collision detection between massive bones and implant models, consisting of millions of triangles at interactive rates on a commodity PC.