人工关节微纳磨粒的多视图空间形态表征与磨损机理研究

Wear debris features are important indicators for wear mechanism analysis and wear performance assessment of artificial joints. However, the micro- and nano-particle feature obtained from a single microscopic view is not able to fully characterize particle spatial morphology. The accuracy in characterizing artificial joint wear mechanism is therefore limited. In this project, we propose to tackle a key scientific problem in complete characterization of the multi-view spatial morphology of artificial joint micro- and nano-scale wear debris and the accurate identification of the wear mechanism. The micro-nano particles are first segmented from multi-view images and their features are extracted using image processing techniques. Multi-view three-dimensional (3D) reconstruction of wear particles is then carried out to characterize their spatial morphologies based on planar and spatial feature transformations. Furthermore, spatial features and the distribution of the micro-nano wear debris are evaluated for different artificial joint wear mechanisms. Based on the obtained information, the relationship between particle spatial morphologies and wear mechanisms is investigated by utilizing information fusion theory, and an intelligent identification algorithm for wear mechanisms of artificial joints is proposed. Finally, experiments are conducted on different working conditions and different wear modes to evaluate the performance of the proposed approach on spatial morphology characterization and wear mechanism identification of artificial joints. This research will provide a new theory and technology for spatial morphology characterization and wear mechanism analysis of artificial joint wear debris in micro- and nano-scale. The outcome is very significant for design optimization and performance improvement of artificial joints.

人工关节的磨粒(磨屑)信息是其磨损机理鉴别和磨损性能评估的重要依据，但迄今由单一显微视图获取的微纳磨粒信息无法全面表征其空间形态特征，限制了人工关节磨损机理的精准判定。为此，项目提出研究人工关节微纳磨粒的多视图空间形态完整表征及其磨损机理精准辨识的关键科学问题。通过研究多视角图像处理理论，提出微纳磨粒的多视图分割和特征匹配算法，建立其平面特征与空间特征的转换模型，获得人工关节微纳磨粒的多视图三维重建和空间形态表征方法；探索不同磨损机理产生的人工关节微纳磨粒的空间形态特征及分布，融合智能信息理论，建立磨粒空间形态与其磨损机理映射关系的理论模型，提出人工关节磨损机理智能辨识算法；揭示不同工况和不同磨损模式下人工关节的磨粒空间形态特征和磨损机理关系并进行理论与实验验证。项目研究将为人工关节微纳磨粒的空间形态表征及磨损机理辨识提供新理论和技术支撑，对人工关节的优化设计和性能提升具有重要意义。

磨粒是磨损的直接产物，是人工关节磨损机理鉴别和磨损性能评估的重要信息载体，但从单一显微视图获取的磨粒特征信息无法全面表征其空间形态特征，导致人工关节磨损机理的判定误差。为此，项目开展了人工关节微纳磨粒的多视图空间形态表征及磨损机理辨识方法研究。主要研究内容：(1)研究人工关节微纳磨粒的多视图三维重建和空间形态表征方法；(2)研究磨粒空间形态映射的人工关节磨损机理辨识方法；(3)磨粒空间形态表征与磨损机理辨识结果的实验验证。设计了运动磨粒视频采集系统和多聚焦图像采集系统，研究基于多视图轮廓拟合与多聚焦图像序列的磨粒表面三维重建方法，构建深度学习网络实现了磨粒图像深度估计，提取了磨粒表面的算术平均高度、均方根高度和峰度等三维特征信息；研究基于混合卷积神经网络的磨粒检测与类型识别方法，实现块状、条状、球状和片状等磨粒类型的智能判别，建立了磨粒形态特征与磨料、疲劳和黏着等典型磨损机理的映射关系；采用生物医用镍钛合金材料制备摩擦副，开展摩擦磨损实验，结合离线磨粒分析技术验证磨粒三维特征提取的有效性，通过磨损表面形貌分析，验证基于磨粒形态特征辨识的磨损机理的可靠性。项目研究成果显著，共发表学术论文14篇，其中SCI论文9篇，中文核心期刊论文3篇，会议论文2篇；申请发明专利5项，实用新型专利2项；获第七届全国大学生生物医学工程创新设计竞赛三等奖1项；培养硕士生5名。