Digital holography in microplastic identification

Plastic debris increase every day, invading the entire ecosystem, especially the marine environment. The damage caused by plastics, and even more by their fragmentation into micrometric samples, called microplastics (MPs), is becoming irrecoverable. Many techniques are used to analyse MPs, but a standardized procedure is still missing. Digital Holography (DH) has proved to be a powerful imaging tool for identifying MPs in water samples, highlighting highthroughput, label-free, high-coherent and non-invasive prowess. Besides, DH furnishes quantitative information, morphological parameters and numerical refocusing, suitable for microfluidic systems. Both physical and chemical information that characterize the optically denser object are completely enclosed in the phase contrast, measured by light waves transition. However, DH is not capable alone to be material specific and to gather polymers information. To overcome these constraints, artificial intelligence (AI) has been considered, demonstrating to be a powerful pawn for accurately identifying MPs samples. Here we identify, characterize, and classify MPs samples by means of DH, empowered by AI, providing an DH modality for a fast and high-throughput analysis of MPs at lab-on-chip scale, distinguishing them from marine diatoms. We use a machine learning (ML) approach on “holographic features” extracted from DH images for distinguishing MPs from diatoms with a well-established SVM classifier. Then, we couple DH microscopy and machine learning to a novel characterization of phase-contrast patterns based on the fractal geometry. Besides, we use a polarization-resolved DH flow cytometer to prove MPs birefringence, and to accurately discriminate between different types of MPs with fiber shape.