Correlative microscpectroscopy for reliable microplastic particle identification

, Brandt Josef, Bittrich Lars, Fischer Franziska, Fischer Dieter, Kanaki Elisavet, Hassellöv Martin, Mattson Karin.

Fast and reliable methods for the identification of microplastic (MP) particles in environmental samples are key to a better understanding of sources, pathways and sinks of MP in the environment. Precise knowledge over particle type and size allows drawing conclusions on the impact on the ecosystems the particles are found in. Automated microspectroscopy techniques, such as Micro Fourier Transform Infrared (FTIR) or Micro Raman spectroscopy are conventionally used to determine number, size and chemical classification of MP particles. Both techniques have their advantages and disadvantages; none can be clearly be put in favor over the other. The self-developed software package GEPARD [1] allows combining different microscopy techniques in a correlative manner to exploit the advantages of each method, whilst minimizing analysis effort and time. With GEPARD, a particle based analysis approach is followed, entailing the following steps: (i) Optical image acquisition, using either the built in cameras in microspectrometers or a dedicated light microscope, (ii) automated image segmentation for particle recognition, (iii) spectroscopic scan of the detected particles with either Raman, or FTIR or also both, (iv) spectrum matching using spectra databases and (v) result reviewing and reporting. GEPARD creates and saves comprehensive datasets for each sample, allowing to measure subset of particles with different techniques. E.g., large particles are measured first with FTIR, then small particles with Raman and finally Raman with a different laser for particles showing strong fluorescence. GEPARD's flexible workflow allows for an optimal utilization of each analysis lab's equipment configuration, while generating consistent data sets that can be analyzed independently of any hardware requirements. This opens up novel pathways of data mining (e.g., using machine learning) to maximize the level of information gained from valuable samples. [1] Brandt, J. et al. doi: 10.1177/0003702820932926.

View online