Rapid identification of pathogens by using surface-enhanced Raman spectroscopy and multi-scale convolutional neural network
| Autor: | Yaoguang Zhong, Jianhua Zhang, Glenn M. Young, Qingqing Lin, Jingyu Ding, Jiameng Zhang, Chun Jiang |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
Serotype
Salmonella Materials science Time Factors Salmonella enteritidis Metal Nanoparticles Surface-enhanced Raman spectroscopy medicine.disease_cause Spectrum Analysis Raman Biochemistry Convolutional neural network Salmonella Paratyphi Analytical Chemistry Rapid identification Calculated data Salmonella Infections medicine Humans Gold Neural Networks Computer Biological system |
| Zdroj: | Analytical and bioanalytical chemistry. 413(14) |
| ISSN: | 1618-2650 |
| Popis: | Salmonella is a prevalent pathogen causing serious morbidity and mortality worldwide. There are over 2600 serovars of Salmonella. Among them, Salmonella Enteritidis, Salmonella Typhimurium, and Salmonella Paratyphi were reported to be the most common foodborne pathogenic serovars in the EU and China. In order to provide a more efficient approach to detect and distinguish these serovars, a new analytical method was developed by combining surface-enhanced Raman spectroscopy (SERS) with multi-scale convolutional neural network (CNN). We prepared 34-nm gold nanoparticles (AuNPs) as the label-free Raman substrate, measured 1854 SERS spectra of these three Salmonella serovars, and then proposed a multi-scale CNN model with three parallel CNNs to achieve multi-dimensional extraction of SERS spectral features. We observed the impact of the number of iterations and training samples on the recognition accuracy by changing the ratio of the number of the training and testing sets. By comparing the calculated data with experimental one, it was shown that our model could reach recognition accuracy more than 97%. These results indicate that it was not only feasible to combine SERS spectroscopy with multi-scale CNN for Salmonella serotype identification, but also for other pathogen species and serovar identifications. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
| Nepřihlášeným uživatelům se plný text nezobrazuje | K zobrazení výsledku je třeba se přihlásit. |