Rapid Machine Learning-Driven Detection of Pesticides and Dyes Using Raman Spectroscopy

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Machine Learning

Background:

• Pesticide and synthetic dye contamination poses significant risks to food safety, human health, and the environment.
• Raman spectroscopy provides molecular fingerprints but faces challenges like spectral noise and fluorescence, hindering practical application.
• Developing rapid, reliable detection methods is crucial for effective monitoring.

Purpose of the Study:

• To develop a robust deep learning framework (MLRaman) for detecting pesticides and dyes using Raman spectroscopy.
• To improve the accuracy and applicability of Raman spectroscopy for contaminant analysis.
• To create a user-friendly tool for real-time monitoring of food and environmental samples.

Main Methods:

• Implemented a deep learning framework utilizing ResNet-18 for feature extraction from Raman spectra.
• Employed advanced classifiers including XGBoost, Support Vector Machines (SVM), and their hybrid integration.
• Utilized dimensionality reduction techniques (PCA, t-SNE, UMAP) for spectral data analysis.
• Developed a Streamlit application for real-time prediction and validation.

Main Results:

• The MLRaman framework, particularly the CNN-XGBoost model, achieved 97.4% predictive accuracy and an AUC of 1.0.
• The CNN-SVM model demonstrated competitive performance with strong class-wise discrimination.
• Dimensionality reduction confirmed clear separability of Raman embeddings for 10 analytes (7 pesticides, 3 dyes).
• The Streamlit application successfully identified unseen spectra from independent experiments and literature, demonstrating strong generalization.

Conclusions:

• Established a scalable and practical MLRaman model for multiresidue contaminant monitoring.
• The developed framework significantly enhances the capabilities of Raman spectroscopy for detecting pesticides and dyes.
• The MLRaman model shows substantial potential for deployment in food safety and environmental surveillance systems.