Giardino Matteo, Balestra Valentina, Janner Davide, Bellopede Rossana
Department of Applied Science and Technology (DISAT), Politecnico di Torino, Corso Duca degli Abruzzi 24, 10129 Torino, Italy; INSTM, Consorzio Interuniversitario Nazionale per la Scienza e Tecnologia dei Materiali, Via G. Giusti 9, 50121 Florence, Italy.
Department of Environment, Land and Infrastructure Engineering (DIATI), Politecnico di Torino, Corso Duca degli Abruzzi 24, 10129 Torino, Italy.
Sci Total Environ. 2023 Feb 10;859(Pt 2):160036. doi: 10.1016/j.scitotenv.2022.160036. Epub 2022 Nov 12.
Microplastics (MPs) are a heterogeneous group of solid polymers with dimensions <5 mm, which are a widespread contaminant of the environment. Their ubiquitous presence grabbed researchers' attention in the last decade, and the problem of MPs detection and quantification is currently a topic of utmost importance. Most identification and quantification protocols are still based on the visual count, which is an extremely time-consuming and error-prone task due to operator subjectivity. To address such an issue, different software analysis procedures are available, but they mainly rely either on the use of optical microscopy, covering a minimal area for each sample (mm size), or they allow only the identification of the largest particles (>1 mm). Here, a semi-automatic innovative image processing method for quantifying and measuring microplastics on filter membrane substrates is presented and validated, comparing results with data obtained using visual counting performed by an experienced operator. The algorithm was tested with artificially generated microplastic images and samples taken from natural environments. Samples of Borgio Verezzi show cave sediment and Po River water were filtered on a glass filter membrane, and photographs were taken under 365 nm illumination, both without and with Nile Red staining. The proposed image analysis method, implemented in an easy-to-use Python script, was quite accurate and fast (about 10 s/image average processing time), showing an average deviation below 10 %, which is further reduced to about 8 % if the samples are stained with Nile Red.
微塑料（MPs）是一类尺寸小于5毫米的异质固体聚合物，是环境中广泛存在的污染物。在过去十年中，它们无处不在的现象引起了研究人员的关注，目前微塑料的检测和定量问题是一个极其重要的课题。大多数识别和定量方案仍基于目视计数，由于操作人员的主观性，这是一项极其耗时且容易出错的任务。为了解决这一问题，有不同的软件分析程序可用，但它们主要依赖于光学显微镜的使用，每个样品覆盖的面积最小（毫米尺寸），或者只允许识别最大的颗粒（>1毫米）。在此，提出并验证了一种用于定量和测量滤膜基质上微塑料的半自动创新图像处理方法，并将结果与经验丰富的操作人员进行目视计数得到的数据进行比较。该算法用人工生成的微塑料图像和从自然环境中采集的样品进行了测试。博尔焦韦雷齐的样品显示洞穴沉积物和波河水在玻璃滤膜上过滤，并在365纳米光照下拍摄照片，分别有无尼罗红染色。所提出的图像分析方法在一个易于使用的Python脚本中实现，相当准确且快速（平均处理时间约为10秒/图像），平均偏差低于10%，如果样品用尼罗红染色，偏差可进一步降至约8%。
