Tsuchiya Masashi, Kitahashi Tomo, Taira Yosuke, Saito Hitoshi, Oguri Kazumasa, Nakajima Ryota, Lindsay Dhugal J, Fujikura Katsunori
Japan Agency for Marine-Earth Science and Technology (JAMSTEC), Yokosuka, Kanagawa, Japan.
KANSO Technos Co., Ltd., Tokyo, Japan.
PeerJ. 2025 Mar 31;13:e19196. doi: 10.7717/peerj.19196. eCollection 2025.
The distribution and migration processes of microplastics (MPs) in the marine sediments have yet to be fully elucidated. To estimate the contamination levels and distribution patterns, and develop countermeasures, the amount of MPs must be understood. Rapid and efficient processing of numerous samples is also needed to detect and determine MP contamination. However, whatever the sample of interest, MP analysis is time consuming. This is especially the case for deep-sea sediments, where the particle sizes are small and pretreatment processes are complex and time-consuming. To address the need for rapid and efficient detection of MPs, we propose a novel method for automatically identifying and counting Nile Red (NR)-stained sedimentary MP particles captured under a stereoscopic fluorescence microscope. In this study, we demonstrated the utility of the developed system by comparing its recovery rate and analysis time with those of the conventional methods used for manual processing. The developed method can efficiently detect MPs of sizes between 18 and 500 µm and classify them as fibers or grains (or fragments). This means that our method can efficiently detect MPs as small as 100 µm found in deep-sea sediments. The semi-automated MP detection system gave a counting time of 4.2-8.8 s per particle-as the number of particles increases, the analysis time per particle decreases. Similarly, when the number of particles counted using a stereomicroscope and image analysis software was set at 100, the automatic measurement method using a flow cell could measure 50-80% of the total number of particles, depending on the type of MPs. By using artificial particulate and fibrous MPs as training data and combining them with a machine learning system, we were able to build a system that can classify both types with 98% accuracy (100% for fibers and 96% for grains). In natural samples, approximately 150 µm (20-350 µm in range) MPs were detected, and the number was consistent with previous studies. This demonstrates the effectiveness of the method we developed. We established a rapid detection method for the number and form of MPs using a continuous semi-automated method, combining NR staining and artificial intelligence. Although this method does not allow the identification of polymer types, it enables that rapid and reliable quantification of MPs numbers. The new method established in this study is expected to improve the accuracy of information on the distribution, destination, and quantity of MPs. It is also relatively easy to use and can transfer technology in various fields, from citizen science to rapid diagnosis on research vessels in the open ocean.
微塑料（MPs）在海洋沉积物中的分布和迁移过程尚未完全阐明。为了估计污染水平和分布模式并制定应对措施，必须了解微塑料的数量。为了检测和确定微塑料污染，还需要对大量样本进行快速有效的处理。然而，无论感兴趣的样本是什么，微塑料分析都很耗时。对于深海沉积物来说尤其如此，那里的颗粒尺寸小，预处理过程复杂且耗时。为了满足快速高效检测微塑料的需求，我们提出了一种新方法，用于自动识别和计数在立体荧光显微镜下捕获的尼罗红（NR）染色的沉积微塑料颗粒。在本研究中，我们通过将开发系统的回收率和分析时间与用于人工处理的传统方法进行比较，证明了该系统的实用性。所开发的方法能够有效地检测尺寸在18至500微米之间的微塑料，并将它们分类为纤维或颗粒（或碎片）。这意味着我们的方法能够有效地检测到深海沉积物中发现的小至100微米的微塑料。半自动微塑料检测系统每个颗粒的计数时间为4.2 - 8.8秒，随着颗粒数量的增加，每个颗粒的分析时间会减少。同样，当使用立体显微镜和图像分析软件计数的颗粒数量设定为100时，使用流动池的自动测量方法根据微塑料的类型能够测量总数的50 - 80%。通过使用人工颗粒状和纤维状微塑料作为训练数据，并将它们与机器学习系统相结合，我们能够构建一个对两种类型进行分类的准确率达到98%（纤维为100%，颗粒为96%）的系统。在天然样本中，检测到了大约150微米（范围为20 - 350微米）的微塑料，数量与先前的研究一致。这证明了我们开发的方法的有效性。我们建立了一种结合尼罗红染色和人工智能的连续半自动方法，用于快速检测微塑料的数量和形态。虽然这种方法无法识别聚合物类型，但它能够快速可靠地对微塑料数量进行定量。本研究中建立的新方法有望提高关于微塑料分布、归宿和数量信息的准确性。它使用起来也相对容易，并且可以在从公民科学到公海研究船上的快速诊断等各个领域转让技术。
