Bringing standardization into the detection of microplastics in products and the environment

The success story of Purency

Benedikt Hufnagl is co-founder and CTO of Purency GmbH, which was founded in 2020. Thanks to his PhD studies  in the field of chemometrics at the Technical University of Vienna and his years of experience in programming, he is not only able to put himself in the position of the user and identify problems in the market, but also to find solutions quickly.

Benedikt Hufnagl develops product solutions and aligns them with the strategy and vision of Purency. At the same time, he takes an active role as an ISO-delegate where he is working to develop an International standard for microplastic analysis.

The pollution of the environment by microplastics is a topic that receives ever more attention both from scientists and the general public. To better understand the impact of this novel contaminant it is indispensable to quantify the abundance of microplastics in their respective habitats. Therefore, many approaches to monitor microplastics have been proposed.

Plastic_Sea water contaminated by micro plastic_GettyImages-1036767280.jpg

Focal plane array-based µFTIR imaging (Fourier transform infrared) is one such technique that allows to measure microplastics down to a size of approx. 10 µm. Even though the FTIR spectra allow for a chemical identification of the found microplastics based on characteristic vibrational bands, the analysis is far from trivial as datasets can typically contain more that 1 million spectra. Common approaches such as database search require 4 to 48 hours of computation time to analyse a single sample, followed by an extensive auditing by experts due to the low statistical performance of this approach. This bottleneck prevents the large-scale application of µFTIR Imaging in routine analysis and monitoring.

Benedikt’s research deals with the development and application of a machine learning approach which can analyse a sample in less than 10 minutes. His research articles describe the methodology on how machine learning models based on random forests can be deduced and a application study of the source of microplastics in bottled water.

The current machine learning model is integrated into a software for ordinary lab PCs and can already analyse more than 20 different kinds of polymers in matrices such as sea water, fresh water, sediments, soil, compost, and sewage sludge.

Based on his research and together with two other TU alumni and a business economist, Benedikt founded the Start-up Purency GmbH. The solutions offered by the Viennese company enable laboratories to perform scalable analyses of microplastics. The high-quality results minimize the need for manual processing. Thus, the time required for data analysis is reduced from hours to approximately ten minutes. Thereby, Purency GmbH creates the basis for reducing microplastics.

Standards and Innovation

Since 2019 Benedikt participates in international standardization in ISO/TC 61/SC 14/WG 4 "Characterization of plastics leaked into the environment (including microplastics) and quality control criteria of respective methods". There he focuses on developing quality as well as comparability criteria for algorithms currently in use for microplastics detection for the purpose of harmonization.

“The big challenge now is to correctly interpret the data obtained. So far, database search approaches have been used primarily for this. However, these are only suitable to a limited extent for large amounts of data and often require lengthy post-processing. The evaluations are then difficult to compare, and detailed and comprehensible statements are hardly possible ”

Benedikt Hufnagl

co-founder and CTO of Purency GmbH

Different approaches have been followed to detect microplastics. One of them is µFTIR  spectroscopy. In FTIR spectroscopy the sample is illuminated with electromagnetic radiation in the infrared range. The light waves overlap each other in a complex wayand the resultingsignal is measured. This results in a characteristic “fingerprint” of the examined sample.

Purency is therefore taking a new approach for the analysis of the measurment: the data is evaluated with the help of machine learning algorithms. This makes it possible to reliably determine the number, type and size of the particles. Large amounts of data can be analysed in a short time. The spectrometer scans the entire sample, and a wavelength spectrum is recorded for each individual pixel. This results in images with a million spectra and 5 GB in size.

More than 20 types of polymers can be differentiated, particles with a size of only 10 micrometers can be recognized - the limit is only set by the resolution of the measuring device used, not by the data processing method.

The result is a detailed table in which all the types of polymer present are shown according to particle size and number - it is available within about ten minutes. The method runs automatically and does not require time-consuming manual post-processing - the trained staff can concentrate on other tasks. Particular emphasis was placed on ensuring that the results are exactly reproducible - repeated analyses reliably deliver the same results.

The Microplastics Finder can be used in many ways, even for heavily polluted environmental samples, such as sewage sludge. The Microplastics Finder also has a strong impact on harmonization of the respective methods. The common database approach forces the user to choose a large set of parameters, which is one of the main reasons, why microplastics data is still only partly comparable. Contrary to that Purency’s machine learning model is fixed and does not allow any kind of manipulation through the user. Therefore, human bias is removed and results from different operators are identical.

In 2021 Purency has been awarded with the Living Standards Award by Austrian Standards International, the Austrian national standardization body, in the category “Developing Future Technologies”.

Further readings