Chinese Journal of Lasers, Volume. 49, Issue 15, 1507405(2022)
Surface Enhanced Raman Scattering Detection of Four Foodborne Pathogens Using Positively Charged Silver Nanoparticles and Convolutional Neural Networks
Figures & Tables(11)
Fig. 1. 1D convolutional neural network for SERS classification. (a) Residual block; (b) structure of ResNet11
Fig. 3. TEM images of AgNPs＋ binding with pathogens. (a) S.aureus; (b) S.aureus@AgNPs; (c) E.coli; (d) E.coli@AgNPs
Fig. 5. Raman spectra of AgNPs, silicon wafer, four pathogens, and pathogen-AgNPs compounds
Fig. 6. SERS printfinger spectra of 10 measurements of four pathogens. (a) S.aureus; (b) V.parahemolyticus; (c) L.monocytogenes; (d) E.coli
Fig. 8. SERS fingerprint spectra and of Raman spectra of S.aureus solutions under different conditions. (a) SERS fingerprint spectra of S.aureus solutions with high molecular concentration and low molecular concentration; (b) Raman spectra of low molecular concentration S.aureus solution with and without AgNPs＋
Table 1. Brief introduction to the used machine learning classification methods
View tableTable 1. Brief introduction to the used machine learning classification methods
Methods Description Supervised Linear Algorithm Key parameter Logistic Single-layer neural networks, searching for the most appropriate function parameters to minimize the distance between the predict and true value Yes Linear Greedy algorithm & traversal search solver＝‘liblinear’ SVM Searching the hyperplane which maximizes the margin between different classes Yes Indirect non-linear Sequence minimization kernel＝‘linear’ RF The prediction label of a new sample is the mean prediction of many decision trees which have been constructed during training Yes Non-linear Greedy algorithm & traversal search default KNN The prediction label of a new sample is depending on its distance (Manhattan distance, European distance, and Minkowski distance) to all available classes Yes Linear K-dimensional tree metric＝‘euclidean’
Table 2. Band assignment of main Raman peaks of pathogens[29-32]
View tableTable 2. Band assignment of main Raman peaks of pathogens[29-32]
Raman shift /cm－1 Assignment 621－624 Aromatic ring skeletal 651－683 δ(COO-) guanine 720－735 Glycosidic ring, adenine 792－797 υ(CN) Tyr 840－852 υ(C—C) 875 υ(C—C) skeleton protein 952－958 υ(CN), protein 1034－1042 CH group, protein 1085－1093 Lipid, nucleic acid 1128 υ(C—C), nucleic acid 1319－1333 Adenine, polyadennine, DNA 1453－1467 δ(CH2), COO-，lipid 1565－1568 δ(NH, CH), υ(C—C) 1581－1591 Adenine, guanine 1640－1688 Amide
Table 3. Classification accuracies of SERS fingerprint spectra of pathogens by different methods
View tableTable 3. Classification accuracies of SERS fingerprint spectra of pathogens by different methods
Measurement condition Classification accuracy /% SVM Logistic RF KNN ResNet11 720 SERS spectra from 4 pathogens (107 mL－1) 95.83 92.64 85.27 90.27 99.30 100 SERS spectra of S.aureus (103 mL－1) 85.00 78.00 71.00 84.00 98.00
Tools
Get Citation
Copy Citation Text
Yong Yang, Hao Dong, Shu Wang, Yaosuo Sang, Zhigang Li, Long Zhang, Chongwen Wang, Yong Liu. Surface Enhanced Raman Scattering Detection of Four Foodborne Pathogens Using Positively Charged Silver Nanoparticles and Convolutional Neural Networks[J]. Chinese Journal of Lasers, 2022, 49(15): 1507405
Paper Information
Category: Bio-Optical Sensing and Manipulation
Received: Dec. 9, 2021
Accepted: Apr. 1, 2022
Published Online: Aug. 5, 2022
The Author Email: Wang Shu (wangshu@aiofm.ac.com), Zhang Long (zhanglong@aiofm.ac.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20