Integrating Raman Spectroscopy and Machine Learning to Classify Microplastics

In spectroscopic circles, one of the biggest hot topics right now in the space is environmental analysis. Specifically, spectroscopists (and chromatographers) are looking at applying new and conventional analytical techniques to better classify microplastics that end up in the water supply and threaten the aquatic wildlife, plant life, and ecosystems. According to a recent study published in the Journal of Water Process Engineering, a research team from the Manipal Academy of Higher Education examined how integrating Raman spectroscopy with machine learning can classify microplastics in water sources (1).

Microplastics could cause significant damage to the environment and aquatic ecosystems. These microscopic particles can enter water streams through industrial processes and by humans improperly disposing of their plastic waste. The accumulation of microplastics, resulting from the pervasive disposal of a variety of plastic waste, poses a substantial threat to aquatic organisms (1,2). These microscopic particles are readily ingested by marine life, leading to the accumulation of harmful substances and toxic components within their systems and disrupting their biological processes (1,2). Therefore, a need exists to accurately identify and mitigate the presence of MPs in water bodies to help preserve aquatic ecosystems.

Lead author Jijo Lukose and his team explored this issue in depth, looking at ways to improve on traditional methods for classifying microplastics in water sources. This process included improving on traditional methods, such as Fourier transform infrared spectroscopy (FT-IR) and pyrolysis-gas chromatography–mass spectrometry (Py-GC–MS) which have several drawbacks, including low resolution, extended imaging time, and limited particle size analysis (1,3). These limitations have hindered the efficiency and accuracy of detecting and classifying microplastics. However, vibrational spectroscopic techniques, such as Raman spectroscopy, have emerged as a better technique to use for this purpose. Raman spectroscopy is a powerful analytical technique that provides detailed information about specific molecular vibrations, which can be used to identify chemical compounds, including microplastics (1).

When combined with machine learning, because of its robust feature extraction capabilities, analysis of the Raman spectral data is improved. Lukose and his team explored various machine learning techniques in conjunction with Raman spectral features for microplastic investigations. The methodologies discussed in the article include principal component analysis (PCA), K-nearest neighbor (KNN), random forest, and support vector machine (SVM), to name a few (1).

These machine learning techniques offer promising improvements over traditional methods. For instance, PCA reduces the dimensionality of Raman spectra, making it easier to visualize and analyze complex data sets (1). KNN is beneficial for classifying microplastics based on the closest training examples in the feature space (1). Random Forest and SVM provide robust classification capabilities that can handle the diversity of microplastic types (1).

However, the research team also highlighted the obstacles using these approaches for microplastics classification. In particular, they underscored the hurdle of preparing and training data (1). Although deep learning algorithms were demonstrated to be effective for this purpose, they also require substantial amounts of data to function and acquiring extensive Raman spectroscopic data is particularly challenging (1).

Jijo Lukose and his team's research advocates for further exploration and advancement in this field. The utilization of machine learning to assist in the analysis of Raman spectroscopy data holds significant potential for improving the detection and classification of microplastics. This approach not only promises to enhance the resolution and speed of microplastic identification, but it also opens new avenues for research and development in combating plastic pollution.

References

(1) Sunil, M.; Pallikkavaliyaveetil, N.; Mithun, N.; et al. Machine Learning Assisted Raman Spectroscopy: A Viable Approach for the Detection of Microplastics. J. Water Proc. Eng. 2024, 60, 105150. DOI: 10.1016/j.jwpe.2024.105150

(2) Asensio-Montesinos, F.; Ramirez, M. O.; Gonzalez-Leal, J. M.; et al. Characterization of plastic beach litter by Raman spectroscopy in South-western Spain. Sci. Total Environ. 2020, 744, 140890. DOI: 10.1016/j.scitotenv.2020.140890

(3) Ren, L.; Liu, S.; Huang, S.; et al. Identification of Microplastics Using a Convolutional Neural Network Based on Micro-Raman Spectroscopy. Talanta 2023, 260, 124611. DOI: 10.1016/j.talanta.2023.124611