Abstract: As a common medical equipment to fix fractured bones, bone plate is supposed to be closely fit with bones. However, until now, the designs of bone plates are mostly based on clinical experience, or just take a few bone geometries as references, which cannot be tightly fit with bones and does not meet most patients’ requirements. In this paper, an approach to obtain a series of “average bones” from the 3D model set of the bones is proposed, which lays a foundation for serial design of bone plates. The approach first adopts Coherent Point Drift non-rigid registration algorithm and Laplacian-coordinate-based mesh deformation algorithm, re-meshing the bone-data to be topology-consistent, and then process by principal component analysis and cluster analysis. Experiments results show that bone plates, manufactured in accordance with such series of “average bones”, can significantly better fit the broken bones, and meet the requirements of practical application.

Key words: 3D bone model, mesh model uniformity, 3D registration, mesh deformation, cluster analysis

摘要： 接骨板作为一种常见的固定断骨用医疗器械，需要与骨骼密切贴合。然而目前接骨板大多依据临床经验，或者仅参考少数骨骼的几何形状进行设计，不能满足需求。因此对骨骼三维模型数据集进行分析，为接骨板系列化设计提供依据。该方法首先利用一致顶点漂移非刚体配准算法与基于拉普拉斯坐标的网格变形算法，使骨骼数据重网格化成拓扑一致的图形；再通过主成分分析、聚类分析，获得一系列“平均骨骼”。按此方法设计的接骨板，能够更好地满足与断骨的贴合性约束，达到实际应用需求。

关键词: 三维骨骼模型, 网格模型平均化, 三维配准, 网格变形, 聚类分析