New Technique Combines Raman Spectroscopy and AI to Accurately Detect Microplastics in Water

New Technique Combines Raman Spectroscopy and AI to Accurately Detect Microplastics in Water © Evgeny -chronicles-stock.adobe.com

Introduction

Microplastics (MPs)—tiny plastic particles less than 5 millimeters in size—have become pervasive pollutants in water bodies worldwide, from rivers and lakes to oceans. These particles originate from a variety of sources, including the breakdown of larger plastic debris, synthetic fibers from textiles, and microbeads in personal care products. Due to their small size, microplastics can be ingested by aquatic organisms, potentially causing harmful effects that cascade through ecosystems and enter the human food chain (1). MPs become nanoplastics (NPs) and further pollute our environment, and can be ingested in large quantities, even from bottled water (2).

Accurate detection and quantification of microplastics in environmental samples are critical for understanding their distribution, sources, and impacts. Traditional analytical methods often involve labor-intensive sample preparation and microscopy techniques that can be time-consuming and limited in chemical specificity. To address these challenges, researchers have increasingly turned to spectroscopic techniques like Raman spectroscopy, which can provide rapid, non-destructive identification of microplastics by analyzing their molecular vibrations (1).

In the study by Luo and colleagues (1), a novel approach combining Raman spectroscopy with convolutional neural networks (CNN) is developed to enhance microplastic detection in water environments. This method leverages the unique spectral fingerprints captured by Raman spectroscopy and the powerful pattern recognition capabilities of deep learning to enable faster and more accurate classification and quantification of microplastics. The research offers a promising pathway for improving environmental monitoring and advancing our understanding of microplastic pollution (1).

Spectroscopic Analysis of Microplastics

The team led by Luo utilized Raman spectroscopy, a powerful vibrational spectroscopic technique, to analyze the chemical composition of microplastics suspended in water samples. Raman spectroscopy is particularly suited for this task due to its ability to generate distinct spectral fingerprints unique to different polymer types, enabling reliable identification without extensive sample preparation (1). According to tthe authors, "6 different sizes of Polyethylene (PE) MPs were used as detection objects and mixed into 5 different actual water environments. The R² and RMSE of CNN for identifying the concentration of PE solutions could reach 0.9972 and 0.033, respectively" (1).

In this study, the researchers collected water samples containing microplastics and exposed them to a Raman laser source. The scattered light was then measured to obtain spectra, reflecting the molecular vibrations characteristic of various plastic polymers. This spectral data formed the basis for subsequent quantitative analysis (1). The authors reported that, "machine learning models such as random forest (RF) and support vector machine (SVM) were compared, and CNN combined with Raman spectroscopy showed significant advantages in identifying the concentration of MPs" (1).

Leveraging Convolutional Neural Networks for Improved Quantitative Analysis

Recognizing the complexity and variability of microplastic spectra, Luo and colleagues applied a convolutional neural network (CNN), a deep learning algorithm well-known for its pattern recognition capabilities, to process the spectroscopic data. The CNN was trained on a comprehensive dataset of Raman spectra representing PE beads of different sizes in different water matrices (1). Samples were prepared by density separation and vacuum-assisted filtration. The PE solutions were accumulated on the surface of low-speed qualitative filter paper. The results indicated that this sample preparation method allowed effective quantitative analysis (1).

This approach allowed the network to learn subtle spectral differences and improve quantitative accuracy. By automating feature extraction and classification, the CNN model efficiently quantified the PE microplastic particles from background signals (1).

Quantitative Results and Environmental Implications

The combined Raman-CNN method demonstrated robust performance in quantifying microplastic concentrations across varied water samples, including freshwater and marine environments. The authors reported high accuracy rates in microplastic quantification, supporting the potential of this integrated technique for routine environmental monitoring (1).

This advancement offers researchers and environmental agencies a tool that speeds up microplastic analysis and enhances reliability for assessing pollution sources and impacts (1).

Collaborative Research Effort

The research was conducted by a multidisciplinary team comprising Y. Luo, W. Su, M.F. Rabbi, Q. Wan, D. Xu, Z. Wang, S. Liu, X. Xu, and J. Wu. Their affiliations include multiple institutions, highlighting a broad collaborative effort to address the challenge of microplastic pollution through innovative analytical science (1).

By merging Raman spectroscopy’s molecular specificity with convolutional neural networks’ pattern recognition power, Luo and team present a novel quantitative tool for microplastic detection in aquatic environments. This approach holds promise for enhancing environmental monitoring protocols and informing mitigation strategies against microplastic pollution (1).

References
(1) Luo, Y.; Su, W.; Rabbi, M. F.; Wan, Q.; Xu, D.; Wang, Z.; Liu, S.; Xu, X.; Wu, J. Quantitative Analysis of Microplastics in Water Environments Based on Raman Spectroscopy and Convolutional Neural Network. Sci. Total Environ. 2024, 926, 171925. DOI: 10.1016/j.scitotenv.2024.171925

(2) Hroncich, C. Analyzing Nanoplastics: An Interview with Scientists from the Columbia University’s Climate School. Spectroscopy 2024, Jan 22. Available at: https://www.spectroscopyonline.com/view/analyzing-nanoplastics-columbia-university-climate-school (accessed 2025-08-12).