基于机器学习的拉曼光谱生物化学分析检测研究

Research on Raman spectroscopic biochemical analysis and detection based on machine learning

为了提高HBV血清化学分析检测的精度,采用醋酸纤维素膜与银纳米颗粒对血清进行分离,并使用拉曼光谱进行检测,提出了基于GWO的特征波长筛选和灰狼优化算法(ELM)的改进方法。GWO-ELM的HBV血清化学检测分析模型,通过比较全波、主成分分析、连续投影法等,对模型进行模式化辨识,从而确定HBV血清Raman的特征波长选择。考虑到参数设置对ELM模式的性能的影响,运用GWO算法优化选择ELM模型参数,利用GWO-ELM技术建立了一种基于连续投射技术的HBV病毒抗体的特异性波长筛查方法。与PSO-ELM、GA-ELM和ELM相比,基于GWO-ELM的HBV血清检测模型的准确度最高。该方法能有效地改善HBV感染的血清学诊断,为HBV血清检测提供新的方法。

In order to improve the accuracy of HBV serum detection,cellulose acetate membrane and silver nanoparticles were used to separate the serum,and Raman spectroscopy was used to detect the serum.An improved method based on GWO feature wavelength screening and gray wolf optimization algorithm(ELM)was proposed.HBV serum chemical detection analysis model based on GWO-ELM was adopted.Firstly,the model was identified by comparing full wave,principal component analysis,continuous projection method,etc.,so as to determine the characteristic wavelength selection of HBV serum Raman spectrum.Secondly,considering the influence of parameter settings on the performance of ELM mode,GWO algorithm was used to optimize the selection of ELM model parameters,and GWO-ELM technology was used to establish a specific wavelength screening method for HBV antibody based on continuous projection technology.Compared with PSO-ELM,GA-ELM and ELM,the accuracy of HBV serum detection model based on GWO-ELM is the highest.This method can effectively improve the serological diagnosis of HBV infection and provide a new method for HBV serum detection.