Strafella Pierluigi, Giulietti Nicola, Caputo Alessia, Pandarese Giuseppe, Castellini Paolo
CNR IRBIM, Ancona, Italy.
DIISM, Polytechnic University of Marche, Ancona, Italy.
Heliyon. 2024 Oct 29;10(21):e39875. doi: 10.1016/j.heliyon.2024.e39875. eCollection 2024 Nov 15.
Marine organisms have been observed ingesting microplastic particles, with field analyses indicating fibers and fragments as prevalent forms. Current microplastic detection methods are mainly time-consuming, susceptible to cross-contamination, and expensive. Furthermore, these techniques, being disruptive, do not allow for the exact localization of the microplastic in the sample. This study proposes a new approach using Computed Tomography (CT scan) and Artificial Intelligence for the automatic and non-destructive detection of microplastics in fishes and other species based on the combination of several factors, such as density and shape. The advantages of this methodology include accurate identification of plastic localization, a low risk of cross-contamination, rapid processing, automatic tomographic measurement, efficient data processing, cost-effectiveness, and a high cost-benefit ratio. The herein results highlight how artificial intelligence applied to conventional techniques can significantly improve precision and efficiency in microplastic research. Indeed, the semantic segmentation model clearly recognized the presence of 100 % of the plastic particles, both in their location and in their volume, accelerating the identification process and surpassing the limitations of traditional spectral analysis methodologies.
海洋生物被观察到摄入微塑料颗粒，现场分析表明纤维和碎片是常见的形式。目前的微塑料检测方法主要耗时、易受交叉污染且昂贵。此外，这些技术具有破坏性，无法确定微塑料在样品中的准确位置。本研究提出了一种新方法，利用计算机断层扫描（CT扫描）和人工智能，基于密度和形状等多种因素的组合，对鱼类和其他物种中的微塑料进行自动无损检测。该方法的优点包括准确识别塑料的位置、交叉污染风险低、处理速度快、自动断层测量、高效数据处理、成本效益高以及成本效益比高。本文的结果突出了将人工智能应用于传统技术如何能显著提高微塑料研究的精度和效率。事实上，语义分割模型清楚地识别出了所有塑料颗粒的存在，包括其位置和体积，加速了识别过程，超越了传统光谱分析方法的局限性。
