Enhanced spectral signatures with Ag nanoarrays in hyperspectral microscopy for CNN-based microplastics classfication.

Abstract

Microplastics are a pervasive pollutant in aquatic ecosystems, raising critical environmental and public health concerns and driving the need for advanced detection technologies. Microscopic hyperspectral imaging (micro-HSI), known for its ability to simultaneously capture spatial and spectral information, has shown promise in microplastic analysis. However, its widespread application is hindered by limitations such as low signal-to-noise ratios (SNR) and reduced sensitivity to smaller microplastic particles. To address these challenges, this study investigates the use of Ag nanoarrays as reflective substrates for micro-HSI. The localized surface plasmon resonance (LSPR) effect of Ag nanoarrays enhances spectral resolution by suppressing background reflections and isolating microplastic reflection bands from interference. This improvement results in significantly increased SNR and more distinct spectral features. When analyzed using a 3D-2D convolutional neural network (3D-2D CNN), the integration of Ag nanoarrays improved classification accuracy from 90.17% to 98.98%. These enhancements were further validated through Support Vector Machine (SVM) analyses, demonstrating the robustness and reliability of the proposed approach. This study demonstrates the potential of combining Ag nanoarrays with 3D-2D CNN models to enhance micro-HSI performance, offering a novel and effective solution for precise microplastics detection and advancing chemical analysis, environmental monitoring, and related fields.