微塑料（MPs）在所有生态系统中普遍存在，影响野生动物并最终影响人类健康。自然样本的复杂性以及用于分离聚合物的处理方法的非特异性使得利用常规光谱技术对环境监测中的数千个颗粒进行表征变得不切实际。解决这一问题的主要方法是对样本的小部分进行分析或仅测量载玻片上的部分颗粒，即亚取样。本文提出了一种基于量子级联激光红外成像技术对反光Kevley载玻片和金涂层滤膜进行亚取样的策略，该技术是监测微塑料的一种有前景的工具。与大多数难以提出通用亚取样方案的先前方法不同，我们引入了基于样品的亚取样概念。这种方法可以在初步计数载玻片上的总颗粒数后，始终应用于特定样本，并突出显示最佳的亚取样区域。此阶段的误差可作为代理变量，以最小化评估颗粒和微塑料数量时的误差，显著提高大规模微塑料监测的可行性。该方法的预测能力针对纤维和碎片、总颗粒和微塑料的数量进行了测试。此外，评估结果按尺寸和聚合物类型进行了分类。在大多数情况下，参考值包含在预测值的置信区间内（通常在68%的置信区间内），相对误差低于25%。例外情况出现在当载玻片上存在非常稀少（一个或两个）特定尺寸或聚合物类型的颗粒时。该方法与其他系统性的特定策略进行了比较。

Microplastics (MPs) are ubiquitous in all ecosystems, affecting wildlife and, ultimately, human health. The complexity of natural samples plus the unspecificity of their treatments to isolate polymers renders the characterization of thousands of particles impractical for environmental monitoring using conventional spectroscopic techniques. Two primary solutions are to analyze a small fraction of the sample or to measure only a subset of particles present over a holder, known as subsampling. A strategy to subsample reflective Kevley slides and gold-coated filters using quantum-cascade laser-based infrared imaging is proposed here, as this technology is a promising tool for MPs monitoring. In contrast to most previous approaches that struggle to propose general subsampling schemes, we introduce the concept of sample-based subsampling. This can be applied ex-ante always and it highlights the best subsampling areas for a sample after a preliminary assay to count the total number of particles on a holder. The error at this stage acts as a proxy to minimize errors when evaluating the number of particles and MPs, significantly enhancing the feasibility of large-scale MPs monitoring. The predictive ability of the approach was tested for fibres and fragments, for total amounts of particles and MPs. Further, the evaluations were disaggregated by size and polymer type. In most situations the reference values were contained in the confidence intervals of the predicted values (often within the 68 % ones) and relative errors were lower than 25 %. Exceptions occurred when very scarce (one or two) items of a given size or polymer were present on the overall holder. The approach was compared to other systematic ad-hoc strategies.