Deep Learning-Based Spectral Extraction for Improving the Performance of Surface-Enhanced Raman Spectroscopy Analysis on Multiplexed Identification and Quantitation

Abstract

Surface-enhanced Raman spectroscopy (SERS) has been recognized as a promising analytical technique for its capability of providing molecular fingerprint information and avoiding interference of water. Nevertheless, direct SERS detection of complicated samples without pretreatment to achieve the high-efficiency identification and quantitation in a multiplexed way is still a challenge. In this study, a novel spectral extraction neural network (SENN) model was proposed for synchronous SERS detection of each component in mixed solutions using a demonstration sample containing diquat dibromide (DDM), methyl viologen dichloride (MVD), and tetramethylthiuram disulfide (TMTD). A SERS spectra dataset including 3600 spectra of DDM, MVD, TMTD, and their mixtures was first constructed to train the SENN model. After the training step, the cosine similarity of the SENN model can achieve 0.999, 0.997, and 0.994 for DDM, MVD, and TMTD, respectively, which means that the spectra extracted from the mixture are highly consistent with those collected by the SERS experiment of the corresponding pure samples. Furthermore, a convolutional neural network model for quantitative analysis is combined with the SENN, which can simultaneously and rapidly realize the qualitative and quantitative SERS analysis of mixture solutions with lower than 8.8% relative standard deviation. The result demonstrates that the proposed strategy has great potential in improving SERS analysis in environmental monitoring, food safety, and so on.

Deep Learning–Assisted Surface-Enhanced Raman Scattering for Rapid Bacterial Identification

“…24 Examples include convolutional neural network (CNN) for SERS-based bacteria identification, 14 PCA combined with K-nearest neighbor (PCA-KNN), and residual network (ResNet) to identify SERS spectra of tumor cells, red blood cells (RBCs), and white blood cells (WBCs). 16 Furthermore, the spectral extraction neural network (SENN) was developed to extract pure molecular spectra from SERS spectra of mixtures, 25 demonstrating better classification capability of DL networks in complex clinical samples. Similar to CNN, which was previously applied to natural language processing, Vision Transformer (ViT) is one of the latest DL methods.…”

“…In this study, we sought to identify bacterial Gram type, species, and antibiotic resistance of specific strains using DL-assisted SERS spectral analysis (SERS-DL). As clinical SERS samples are of limited availability, the massive amount of bacterial data required for ML methods were usually obtained from laboratory cultivation or various data augmentation methods such as translation or scaling, which may introduce data bias and not accurately represent the variability of bacteria in the real world. , To address this issue and overcome the limitation of previous DL-based bacterial identification studies, we adopted a total of 11,774 SERS spectra obtained from bacteria directly isolated from clinical blood samples to train our SERS-DL model. Using this approach, we create a more comprehensive and representative training dataset that better reflects the real clinical conditions.…”

Multiplex SERS detection of polycyclic aromatic hydrocarbon (PAH) pollutants in water samples using gold nanostars and machine learning analysis

“…CNN has been used to analyze SERS spectra for the identification of pathogenic bacteria, 55 and for the identification and quantification of multiplexed pesticides. 56 We previously reported CNN as the most effective model for SERS analysis of biomarkers for medical diagnostics. 54 In this study, we demonstrated the first application of CNN on SERS spectra for classification and quantification of PAHs.…”

Multicomponent Raman spectral regression using complete and incomplete models and convolutional neural networks

“…Further, prior studies also often neglect to compare CNN performance to state-of-the-art “traditional” methods. 32…”

“…Further, prior studies also often neglect to compare CNN performance to state-of-the-art "traditional" methods. 32 To address this challenge, we propose a multi-output mixed Raman spectra concentration regression network based on CNN for simultaneous and accurate concentration predictions. To solve the practical constraint of limiting the number of Raman spectra required for training, we further propose a general scheme for generating unlimited synthetic datasets of mixtures which can be used for both Raman classification and quantitative prediction applications.…”