Hyperspectral imaging assists in identifying plastics from medical masks

A paper by KFU’s Bionanotechnology Lab saw light in Analyst.

The COVID pandemic has enormously increased the number of used disposable masks in the environment. They pose a danger of plastic pollution.

“At the first stage, our group collected spectral profiles of solid polypropylene layers of medical masks, which were then used to train a neural network model. After that, the mask samples were subjected to ultraviolet aging, as a result of which hyperspectral data of recycled microplastic particles was obtained. To determine the relationship between the spectra extracted from whole layers and the spectra of microparticles detected during the degradation of masks, a one-dimensional convolutional neural network model with 1.8 million parameters was used,” says co-author, PhD student Ilnur Ishmukhametov.

The neural network showed an 81% accuracy rate in detecting microplastics in specimens.

“The decrease in the accuracy of identification of microplastic particles compared to the whole layers of the mask indicates possible structural changes caused by ultraviolet radiation. However, despite this, the application of the deep learning algorithm in hyperspectral analysis has surpassed the traditional method of spectral angular mapping in the classification of intact and UV-irradiated samples, which confirms the potential of the proposed approach in the analysis of recycled microplastics,” concludes Ishmukhametov.