Raman Spectroscopy Enhanced By Machine Learning For Effective Microplastic Detection In Aquatic Systems

Abstract

Microplastics (MPs), small plastic fragments ranging from 1 µm to 5 mm, pose a growing threat to aquatic ecosystems and human health due to their persistence, toxicity, and ability to bioaccumulate. Conventional methods for identifying MPs are often limited by their dependence on labor-intensive procedures, long analysis times, and sensitivity to environmental interference. Raman spectroscopy (RS), known for its non-destructive nature and molecular specificity, has emerged as a promising technique for MP detection. However, standalone RS suffers from challenges such as weak signal intensity, spectral noise, and manual interpretation constraints. This study explores the integration of RS with machine learning (ML) techniques—including Random Forest, Support Vector Machine, Multilayer Perceptron, k-Nearest Neighbors, and deep learning models such as Convolutional Neural Networks (CNNs) and Autoencoders—to improve MP classification and analysis. The results indicate that ML-assisted RS significantly enhances detection accuracy, reduces reliance on manual analysis, and supports high-throughput, real-time environmental monitoring. Notably, CNN-based models achieved classification accuracies exceeding 99%, even in complex matrices and low signal-to-noise conditions. This hybrid approach demonstrates strong potential for scalable and precise microplastic detection across various environmental domains.