机器学习在医用金属材料特性研究中的应用  

Application of machine learning in key properties of medical metal materials

作　　者：史榴 梁鹏晨 常庆 宋二红 Shi Liu;Liang Pengchen;Chang Qing;Song Erhong(College of Microelectronics,Shanghai University,Shanghai 201800,China;Shanghai Key Laboratory of Gastric Neoplasms,Shanghai Institute of Digestive Surgery,Ruijin Hospital,Shanghai Jiao Tong University School of Medicine,Shanghai 200020,China;State Key Laboratory of High-Performance Ceramics and Ultrastructure,Shanghai Institute of Ceramics,Chinese Academy of Sciences,Shanghai 200050,China)

机构地区：[1]上海大学微电子学院,上海市201800 [2]上海交通大学医学院附属瑞金医院消化外科研究所,上海市胃肿瘤重点实验室,上海市200020 [3]中国科学院上海硅酸盐研究所高性能陶瓷和超微结构国家重点实验室,上海市200050

出　　处：《中国组织工程研究》2024年第17期2766-2773,共8页Chinese Journal of Tissue Engineering Research

基　　金：国家自然科学基金资助项目(81670968),项目负责人:常庆。

摘　　要：背景:机器学习与医用金属材料的结合,弥补传统实验和计算模拟的低效性和高成本的不足,通过分析大量数据快速准确地预测金属材料特性,优化材料设计和性能,提高医学应用的安全性和效率。目的:总结并归纳机器学习在医用材料特性中的研究进展及不足。方法:由第一作者通过计算机检索中国知网、PubMed、X-MOL和Web of Science数据库2013年1月至2023年4月的相关文章。中文检索词为“医用金属材料机器学习,医用钛合金,医用镁合金,医用金属材料性能”,英文检索词为“machine learning medical metal materials,medical stainless steel alloy,medical cobalt-chromium alloy,medical titanium alloy,medical magnesium alloy”,最终纳入70篇相关文献进行归纳总结。结果与结论:①随着传统实验和计算模拟方法所产生的大量数据的可获取性提高,机器学习作为材料设计方法的引入为材料科学研究开辟了新的范式。②机器学习工作流主要分为4个部分:数据收集及预处理、特征工程、模型选择及训练和模型评估,每个环节不可缺少。③医用金属材料分为:不锈钢共基合金、钴铬合金、钛合金和镁合金。针对不锈钢共基合金,机器学习预测其力学性能,要提高机器学习的泛化能力;针对钴铬合金,机器学习预测其力学性能,可得出钴铬合金为髋关节植入物的最佳材料;针对钛合金,机器学习预测其力学性能,可选择出力学性能最优异的植入物;针对镁合金,机器学习预测其耐腐蚀性和力学性能,集成模型可准确预测镁合金的力学性能,随机森林模型可预测镁合金作为血管支架时的最优元素含量。④机器学习在医用材料领域存在一定局限性,如模型相对滞后、数据未能标准化及泛化性较低;未来研究解决此类问题应充分利用深度学习和分割算法技术,使用统一标准数据,改善模型提高泛化能力。BACKGROUND:The combination of machine learning and medical metal materials can make up for the inefficiency and high cost of traditional experiments and computational simulations,and quickly and accurately predict the characteristics of metal materials by analyzing large amounts of data,optimize material design and performance,and improve the safety and efficiency of medical applications.OBJECTIVE:To summarize the research progress and shortcomings of machine learning in the characteristics of medical materials.METHODS:The first author searched CNKI,PubMed,X-MOL,and Web of Science databases by computer to search all relevant articles from January 2013 to April 2023.The Chinese search terms were“machine learning of medical metal materials,medical titanium alloy,medical magnesium alloy,medical metal material properties”.The English search terms were“machine learning medical metal materials,medical stainless steel alloy,medical cobalt-chromium alloy,medical titanium alloy,medical magnesium alloy”.Finally,70 relevant articles were included for a summary.RESULTS AND CONCLUSION:(1)The introduction of machine learning as a material design methodology has opened up new paradigms for material science research as the accessibility of large amounts of data generated by traditional experimental and computational simulation methods increases.(2)The machine learning workflow is divided into four main parts:data collection and preprocessing,feature engineering,model selection and training,and model evaluation,each of which is indispensable.(3)Medical metal materials are categorized into:stainless steel co-base alloys,cobalt-chromium alloys,titanium alloys,and magnesium alloys.For stainless steel co-base alloy,machine learning predicts its mechanical properties,to improve the generalization ability of machine learning.For cobalt-chromium alloy,machine learning predicts its mechanical properties,and it can conclude that cobalt-chromium alloy is the optimal material for hip implants.For titanium alloy,machine learning pred

关 键 词：医用金属材料 机器学习 材料特性 腐蚀性能 力学性能 不锈钢共基合金 钴铬合金 钛合金 镁合金 特征参数 

分 类 号：R459.9[医药卫生—治疗学] R318[医药卫生—临床医学] TB181[一般工业技术]