We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Automated Classification of Undegraded and Aged Polyethylene Terephthalate Microplastics from ATR-FTIR Spectroscopy using Machine Learning Algorithms
Summary
Researchers applied machine learning algorithms to infrared spectroscopy data and achieved 99% accuracy in classifying both fresh and weathered PET microplastics, suggesting that automated tools could greatly speed up the detection and monitoring of plastic pollution in environmental samples.
Automated analysis of microplastics is essential due to the labor-intensive, time-consuming, and error-prone nature of manual methods. Attenuated Total Reflectance Fourier Transform Infrared (ATR-FTIR) spectroscopy offers valuable molecular information about microplastic composition. However, efficient data analysis tools are required to effectively differentiate between various types of microplastics due to the large volume of spectral data generated by ATR-FTIR. In this study, we propose a machine learning (ML) approach utilizing ATR-FTIR spectroscopy data for accurate and efficient classification of undegraded and aged polyethylene terephthalate (PET) microplastics (MPs). We evaluate seven ML algorithms, including Random Forest (RF), Gradient Boosting (GB), Decision Tree (DT), k-Nearest Neighbors (k-NN), Logistic Regression (LR), Support Vector Machine (SVM), and Multi-Layer Perceptron (MLP), to assess their performance. The models were optimized using fivefold cross-validation and evaluated using multiple metrics such as confusion matrix, accuracy, precision, recall (sensitivity), and F1-score. The experimental results demonstrate exceptional performance by RF, GB, DT, and k-NN models, achieving an accuracy of 99% in correctly classifying undegraded and aged PET MPs. The proposed approach capitalizes on the potential of ATR-FTIR spectra to discern distinct chemical signatures of undegraded and aged PET particles, enabling precise and reliable classification. Furthermore, the method offers the benefit of automating the classification process, streamlining the analysis of environmental samples. It also presents the advantage of providing an effective means for method standardization, facilitating more automated and optimized extraction of information from spectral data. The method’s versatility and potential for large-scale application make it a valuable contribution to the field of MP environmental research.