Machine Learning-Assisted Multicolor Fluorescence Assay for Visual Data Acquisition and Intelligent Inspection of...

Area of Science:

• Analytical Chemistry
• Food Science
• Biotechnology

Background:

• Pesticide residues in food pose significant health risks.
• Current detection methods lack efficiency, stability, and struggle with mixed pesticide analysis.
• Innovative techniques are crucial for reliable food safety assessment.

Purpose of the Study:

• To develop an intelligent food risk evaluation system for identifying and quantifying multiple pesticide residues.
• To overcome limitations of existing sensors in terms of efficiency, stability, and mixture detection.
• To integrate advanced analytical methods with machine learning for enhanced pesticide analysis.

Main Methods:

• Utilized a multicolor fluorescent responsive assay with carbon dots (CDs) for pesticide interaction.
• Employed machine learning (ML) algorithms to analyze color changes and extract feature values.
• Implemented a dual-source data acquisition strategy for robust detection, minimizing matrix interference.
• Developed a "stepwise prediction" strategy for automated qualitative and quantitative analysis.

Main Results:

• Achieved 99.3% accuracy in qualitative pesticide identification.
• Obtained high quantitative prediction accuracy (R² ≥ 0.8946) for pesticide concentration.
• Demonstrated effectiveness across six different food matrices.
• Significantly improved detection stability and efficiency compared to traditional methods.

Conclusions:

• The developed intelligent system offers a promising tool for effective food safety monitoring.
• The integration of fluorescent assays and ML provides a powerful approach for complex pesticide residue analysis.
• This method enhances the ability to identify and quantify multiple pesticides, ensuring safer food products.