文献类型：期刊文献

英文题名：Deep Learning-Powered Dark-Field Microscopy for Simultaneous Size and Concentration Analysis of Nanoplastics in Water

作者：Wang, Yi[1];Xi, Cheng Ye[1];Yu, Jun Jie[1];Shao, Yi Ni[1];Wu, Da Jun[1];Qian, Ruo Can[1,2];Chen, Bin Bin[1,2];Li, Da Wei[1,2]

机构：[1]East China Univ Sci & Technol, Sch Chem & Mol Engn, Shanghai 200237, Peoples R China;[2]East China Univ Sci & Technol, Feringa Nobel Prize Scientist Joint Res Ctr, Frontiers Sci Ctr Materiobiol & Dynam Chem, Key Lab Adv Mat,Shanghai Key Lab Funct Mat Chem, Shanghai 200237, Peoples R China

年份：2026

卷号：98

期号：1

起止页码：204

外文期刊名：ANALYTICAL CHEMISTRY

收录：;EI(收录号：20260319906198);WOS:【SCI-EXPANDED(收录号:WOS:001648551100001)】；

基金：This work was supported by the National Natural Science Foundation of China (22176058, 22504037, 22574050), the Science and Technology Commission of Shanghai Municipality (24DX1400200, 23ZR1416100, 25ZR1401082), the Program of Introducing Talents of Discipline to Universities (B16017), and the Fundamental Research Funds for the Central Universities (222201717003). We thank the Research Center of Analysis and Test of East China University of Science and Technology for help with the characterization.

语种：英文

外文关键词：Convolutional neural networks - Image processing

摘要：Nanoplastics have become a significant environmental and health concern due to their widespread presence. Accurately analyzing both size and concentration of nanoplastics is essential for assessing their environmental behavior and potential toxicity; however, this remains a significant challenge. In this study, we developed a novel approach of convolutional neural networks (CNNs) powered dark-field microscopy (DFM) to achieve concurrent size and concentration analysis of nanoplastics. DFM images of polystyrene nanoplastics (PSNPs) down to 150 nm were facilely acquired based on their scattering characteristics, which were subsequently extracted and studied by combining contour recognition algorithms with a streamlined VGGNet. The established approach achieves high accuracy (over 0.99 on test sets) and sensitivity (limit of detection: 1.7 ng mL-1) in identifying PSNPs ranging from 150 to 600 nm. Spiked recovery results yield 93.55-103.8% recovery rates across 200 to 400 nm PSNPs, demonstrating the ability of the developed method to simultaneously determine size and concentration of nanoplastics. Therefore, the proposed strategy can offer a reliable and visual alternative for nanoplastics analysis with potential applications in environmental and biological monitoring.