基于可视化仿生鼻的酮体标志物检测研究

摘要

The existing detecting methods can not fully meet the demand to achieve noninvasive,intuitive and easy detecting of ketone body level of potential ketosis patients outside the hospital or the laboratory.Considering the advantages of high selectivity,high characterization,being noninvasive,intuition,visualization,etc. Experimental research on detection of acetone gas which is the marker of ketone body based on colorimetric artificial nose is carried out,dynamic response analysis,hierarchical clustering analysis,BP neural network identification,etc,are carried out on hue saturation intensity(HSI)color difference characteristic obtained by experimental results.The results show that there is a nonlinear relationship between HSI difference characteristics of visualization bionic nose and the concentration of acetone gas,and BP neural network can identify acetone gas of different concentrations through the HSI difference,overall average relative error of recognition results is 6.67%.%现有的酮体水平检测方法无法完全满足潜在酮症患者在医院外或实验室外的无创、直观简便检测需求.考虑到高选择性、高特征性、无创、直观可视化等优点,基于可视化仿生鼻对酮体标志物丙酮气体的检测进行了实验研究,并对实验结果中所获取的HSI颜色差值特征进行了动态响应分析、层次聚类分析、反向传播(BP)神经网络识别等.分析结果表明:可视化仿生鼻的HSI差值特征与丙酮气体浓度存在非线性关系,且BP神经网络能够通过HSI差值特征实现不同浓度丙酮气体的识别,识别结果的整体平均相对误差为6.67%.