Reference database design for the automated analysis of microplastic samples based on Fourier transform infrared (FTIR) spectroscopy

Abstract

The identification of microplastics becomes increasingly challenging with decreasing particle size and increasing sample heterogeneity. The analysis of microplastic samples by Fourier transform infrared (FTIR) spectroscopy is a versatile, bias-free tool to succeed at this task. In this study, we provide an adaptable reference database, which can be applied to single-particle identification as well as methods like chemical imaging based on FTIR microscopy. The large datasets generated by chemical imaging can be further investigated by automated analysis, which does, however, require a carefully designed database. The novel database design is based on the hierarchical cluster analysis of reference spectra in the spectral range from 3600 to 1250 cm−1. The hereby generated database entries were optimized for the automated analysis software with defined reference datasets. The design was further tested for its customizability with additional entries. The final reference database was extensively tested on reference datasets and environmental samples. Data quality by means of correct particle identification and depiction significantly increased compared to that of previous databases, proving the applicability of the concept and highlighting the importance of this work. Our novel database provides a reference point for data comparison with future and previous microplastic studies that are based on different databases. Graphical abstractᅟ Electronic supplementary materialThe online version of this article (10.1007/s00216-018-1156-x) contains supplementary material, which is available to authorized users.

Microplastic concentrations, size distribution, and polymer types in the surface waters of a northern European lake

“…This result emphasizes the need to examine microliter by spectroscopic or other suitable analytical method, as pointed out previously (Rocha‐Santos & Duarte, ). Recent studies report fully automatic μFTIR spectroscopic analysis with a focal plane array detector (Primpke, Wirth, Lorenz, & Gerdts, ). That procedure could be, when applicable, the most efficient and accurate method for MP determination, if MPs can be satisfyingly isolated and recovered from the sample matrix.…”

Library based identification and characterisation of polymers with nano-FTIR and IR-sSNOM imaging

“…The In order to evaluate the chance of differentiation between polymers within this range, we carried out a hierarchical cluster analysis based on our open access polymer reference library that is applied in the identification of microplastics. 13 The library contains 326 spectra of common commercial polymers and co-polymers as well as natural polymers such as cellulose. Similar to the adaptable database design 13 the Hellinger distance of the spectra was calculated with PRIMER 6 software followed by hierarchical cluster analysis.…”

“…For each sample, three spots were measured with the same settings and the obtained nano-FTIR spectra were averaged. Analogous to established procedures for ATR-IR and µ-FTIR based identification of polymers, the resulting spectrum was then compared to the group's open access database of polymer spectra for automated analysis with the software OPUS (OPUS 7.5, Bruker Optik GmbH) 13. In case of the wider spectral ranges our in-house database for manual identification was applied (the range of the database for automated analysis is 3600-1250 cm 1).…”

A methodology for the fast identification and monitoring of microplastics in environmental samples using random decision forest classifiers

“…For this preliminary study about 100 spectra were sampled for each of the polymer classes by three spectroscopy experts. To ensure enough variability in the matrix and the background 2770 spectra were sampled from both the articially enhanced datasets and from two environmental samples published by Primpke et al 8 which sums up to a total of 3270 spectra. As stated above the validation of the RDF does not require separate test data though for reasons of better comparability to other classication algorithms we further divided each class into a randomly sampled training and test set of equal size.…”

“…Currently the common approach is to perform MP identication in a semiautomatic manner where a spectroscopy expert compares selected pixel spectra to a reference database. 8 At rst glance the obvious solution to speed up this process is to automatically compare each pixel to the database without any human intervention. However, this approach is not only slow but also results in high error rates as current database search routines oen either do not recognize certain MP spectra or falsely assign them to a different type of polymer.…”