Component identification for the SERS spectra of microplastics mixture with convolutional neural network

Area of Science:

• Environmental Science
• Analytical Chemistry
• Spectroscopy

Background:

• Microplastic (MP) pollution is a growing global concern.
• Surface-enhanced Raman spectroscopy (SERS) is a key technique for MP identification due to its unique spectral fingerprints.
• Analyzing complex SERS spectra from MP mixtures remains a significant challenge.

Purpose of the Study:

• To develop an innovative method for simultaneously identifying and analyzing individual components within microplastic mixtures using SERS.
• To evaluate the effectiveness of convolutional neural networks (CNNs) for this analysis.
• To compare CNN performance against traditional algorithms.

Main Methods:

• A convolutional neural network (CNN) model was developed to analyze SERS spectra of six common microplastic mixtures.
• The CNN model was trained using raw, unpreprocessed spectral data.
• Performance was compared to Support Vector Machine (SVM), Principal Component Analysis Linear Discriminant Analysis (PCA-LDA), Partial Least Squares Discriminant Analysis (PLS-DA), Random Forest (RF), and K-Nearest Neighbor (KNN) algorithms.

Main Results:

• The CNN model achieved an average identification accuracy of 99.54% for microplastic components.
• This high accuracy was obtained using raw, unpreprocessed SERS spectral data.
• CNN significantly outperformed all tested traditional algorithms, both with and without spectral preprocessing.

Conclusions:

• Convolutional neural networks offer a highly accurate and efficient solution for identifying microplastic mixtures from SERS spectra.
• The CNN approach eliminates the need for complex spectral preprocessing steps, streamlining analysis.
• This advancement facilitates rapid and precise detection of microplastic pollutants.