A microneedle-based integrated three-electrode system for pesticide detection using machine learning

Pesticides contribute to enhanced agricultural productivity, yet excessive residues pose significant health risks to humans as they persist even after washing, making their detection in crops critically important. In this study, we employed 3D-printed microneedle arrays (MNs) integrated with differential pulse voltammetry (DPV) and a deep learning (DL) algorithm to capture the electrochemical characteristic signals of pesticide molecules. To enhance sensing performance, the working electrode composed of an Au film was further modified with carbon nanotubes, achieving an increase in the surface area and a significantly improved current response. Successful classification and identification were demonstrated on predefined unknown pesticide samples (MS222, <i>p</i>-nitrophenol, crystal violet, malachite green, vanillin, and nitrofurazone) through electrochemical fingerprint analysis. The experimental results revealed that all algorithms attained average accuracies exceeding 90% in interpreting DPV fingerprints, with the convolutional neural network (CNN) attaining 100% classification accuracy, thereby confirming the method's efficacy in pesticide discrimination. In addition, the performance on the extended dataset is also satisfactory. This innovative integration of DPV and DL technologies paves a novel pathway for pesticide classification and recognition, offering substantial potential to advance agricultural safety protocols and safeguard public health.