Related Experiment Video
Updated: Mar 28, 2026
09:57
Multiplex Chemical Imaging Based on Broadband Stimulated Raman Scattering Microscopy
Published on: July 25, 2022
4.7K
Multicomponent Simultaneous Identification Network (MSINet): An Advanced Deep Learning Model for Boosting Multiplex
Xuehua Zhang1, Hansheng Li2, Hao Si1
1School of Chemistry, Chemical Engineering and Life Sciences, Wuhan University of Technology 430070, Wuhan, China.
Analytical Chemistry
|March 27, 2026
Summary
A new deep learning model, MSINet, enhances surface-enhanced Raman spectroscopy (SERS) for analyzing untreated samples. This advanced method accurately identifies and quantifies multiple targets in complex mixtures, overcoming previous detection challenges.
Area of Science:
- Analytical Chemistry
- Spectroscopy
- Machine Learning
Background:
- Surface-enhanced Raman spectroscopy (SERS) offers high sensitivity and specificity for on-site analysis of diverse sample types.
- Challenges in SERS analysis include matrix complexity, competitive adsorption, signal fluctuations, and spectral overlap, hindering multi-target detection in real-world samples.
Purpose of the Study:
- To develop a deep learning model for simultaneous identification and quantification of multiple targets in untreated samples using SERS.
- To create a robust and accurate method that overcomes spectral interferences and signal variability inherent in complex matrices.
Main Methods:
- Designed and implemented a deep learning model, the multicomponent simultaneous identification network (MSINet), for SERS spectral analysis.
- MSINet employs hierarchical feature extraction and multiple feature integration to enhance robustness against signal fluctuations and spectral overlaps.
- A threshold judgment rule was incorporated for efficient quantification of target contents with minimal training data.
Main Results:
- MSINet achieved high accuracy (≥0.90) in detecting and quantifying multiple UV absorbers in seawater, food additives in cocktails, and biomarkers in human serum.
- The model demonstrated significant improvement over the standard curve method, which yielded accuracies as low as 0.05 for the same untreated samples.
- MSINet proved robust against signal fluctuations and spectral overlaps, enabling reliable multi-component analysis.
Conclusions:
- MSINet is a powerful and universal tool for assisting SERS sensors in high-throughput on-site detection of multiple targets in untreated real samples.
- The model's ease of use, graphical interface, and zero extra cost position it for broad application in various analytical fields.
- This deep learning approach significantly advances the capabilities of SERS for complex sample analysis and on-site monitoring.
