Identification of microplastics using a convolutional neural network based on micro-Raman spectroscopy.

Area of Science:

• Environmental Science
• Analytical Chemistry
• Data Science

Background:

• Microplastics (MPs) are a significant global environmental concern.
• Current methods for identifying and classifying MPs are often slow and unreliable.
• Developing rapid and accurate MP analysis techniques is crucial.

Purpose of the Study:

• To develop an efficient and automatic method for microplastic identification.
• To assess the impact of environmental stressors on microplastics.
• To establish robust models for classifying MPs in environmental samples.

Main Methods:

• Utilized micro-Raman spectroscopy for spectral data acquisition.
• Developed a convolutional neural network (CNN) for MP classification.
• Constructed an interaction network based on Raman band correlations to analyze environmental stress effects.

Main Results:

• The CNN model achieved high average classification accuracies: 96.43% for reference MPs and 95.6% for environmental samples.
• An interaction network effectively characterized spectral changes in MPs due to environmental stressors.
• Demonstrated sensitive, information-dense signatures for MP environmental exposure.

Conclusions:

• The combined micro-Raman spectroscopy and CNN approach provides efficient and accurate microplastic identification.
• The interaction network offers a powerful tool for understanding MP degradation and transformation in the environment.
• This study facilitates automatic analysis and intuitive visualization of spectral changes in microplastics.