Intelligent Electrochemical Sensing: Machine Learning-Powered Multidimensional Fingerprinting for Simultaneous...

Area of Science:

• Analytical Chemistry
• Electrochemistry
• Machine Learning

Background:

• Antibiotic residues in food, environmental, and biological systems pose a global public health risk.
• Accurate detection of multiple, structurally similar antibiotics in complex matrices is challenging due to signal overlap.

Purpose of the Study:

• To develop an intelligent electrochemical strategy for simultaneous detection of six antibiotics.
• To address the challenge of signal overlap in multicomponent antibiotic analysis.
• To provide a rapid and accurate method for detecting trace pollutants in complex samples.

Main Methods:

• A single-electrode four-channel electrochemical platform was developed.
• Glassy carbon electrode modified with poly(p-aminobenzenesulfonic acid), HKUST-1, and Au-Pt nanoparticles.
• Square wave voltammetry (SWV) with modulated scan direction and pH for multidimensional signal acquisition.
• Machine learning algorithms (MLP and CNN) for signal decoding, classification, and quantification.

Main Results:

• High classification accuracy (99.75%) for six antibiotics using MLP.
• Excellent regression performance (R² > 0.9990) for concentration prediction using CNN.
• The CNN model demonstrated high accuracy (R² > 0.9850) in real samples (milk, water, serum), with prediction accuracy in serum > 0.9964.
• Achieved picomolar detection limits for antibiotics.

Conclusions:

• An intelligent electrochemical sensing paradigm integrating electrode engineering, signal acquisition, and AI was established.
• The developed method offers a promising strategy for rapid and accurate detection of multiple trace pollutants.
• This approach advances the field of electrochemical sensing for complex environmental and biological monitoring.