面向生物医学检测的拉曼光谱图像机器学习算法研究  

Research on Machine Learning Algorithms for Raman Spectroscopy Imaging for Biomedical Detection

作　　者：于铠铭 包晓栋 李备 洪喜[5] 刘景鑫[3] YU Kaiming;BAO Xiaodong;Li Bei;HONG Xi;LIU Jingxin(Department of Hand Surgery,China-Japan Union Hospital of Jilin University,Changchun Jilin 130033,China;Medical Imaging Engineering Center,China-Japan Union Hospital of Jilin University,Changchun Jilin 130033,China;Department of Radiology,China-Japan Union Hospital of Jilin University,Changchun Jilin 130033,China;HOOKE Instruments Ltd.,Changchun Jilin 130033,China;Changchun Institute of Optics,Fine Mechanics and Physics,CAS,Changchun Jilin 130033,China)

机构地区：[1]吉林大学中日联谊医院手外科,吉林长春130033 [2]吉林大学中日联谊医院医学影像工程中心,吉林长春130033 [3]吉林大学中日联谊医院放射科,吉林长春130033 [4]长春长光辰英生物科学仪器有限公司,吉林长春130033 [5]中国科学院长春光学精密机械与物理研究所,吉林长春130033

出　　处：《中国医疗设备》2021年第8期26-29,共4页China Medical Devices

基　　金：国家重点研发计划(2018YFC1315604;2018YFC0116900);吉林省科技发展计划项目(20200901017SF);吉林大学高层次科技创新团队建设项目(2017TD-27)。

摘　　要：目的探讨拉曼光谱图像进行快速生物医学检测的机器学习处理分析算法,为建立便捷快速的乙肝创新检测方法提供参考。方法使用t-SNE聚类算法和KNN分类算法,对拉曼光谱生物医学检测数据进行处理和分析,验证实验中应用拉曼光谱仪采集乙肝感染血清及正常人血清样本的拉曼光谱数据,通过机器学习算法对拉曼光谱数据进行处理分析,验证算法对拉曼光谱实验数据处理的有效性。结果利用t-SNE聚类算法和KNN分类算法进行拉曼光谱数据处理后,可以有效区分乙肝感染患者血清与对照的正常人血清。结论利用拉曼光谱光谱仪采集生物医学样本光谱图像数据,通过t-SNE和KNN等机器学习算法进行处理分析,是一种可行的快速生物医学检测新方法。Objective The machine learning processing and analysis algorithms for Raman spectral images for rapid biomedical detection are discussed to provide a reference for the establishment of a convenient and fast innovative detection method for hepatitis B.Methods Processing and analysis of Raman spectroscopy biomedical assay data using t-SNE clustering algorithm and KNN classification algorithm.Validation experiments applied Raman spectroscopy to collect Raman spectroscopy data from hepatitis B infected serum and normal human serum samples,and machine learning algorithms were used to process and analyze the Raman spectroscopy data to verify the effectiveness of the algorithms on Raman spectroscopy experimental data processing.Results Raman spectral data processing using t-SNE clustering algorithm and KNN classification algorithm can effectively distinguish hepatitis B-infected sera from control normal human serum.Conclusion The collection of biomedical sample spectroscopy imaging data using Raman spectroscopy and processing and analysis by machine learning algorithms such as t-SNE and KNN is a viable new method for rapid biomedical detection.

关 键 词：生物医学检测 拉曼光谱 机器学习 t分布随机近邻嵌入 K最近邻算法 

分 类 号：R318[医药卫生—生物医学工程]