Ultra-sensitive and specific detection of clenbuterol hydrochloride via Gramian angular field encoding and Conformer-based electrochemical sensing.

Sensitive and specific detection of clenbuterol hydrochloride (CLB) is crucial for ensuring food safety and effective doping control. An electrochemical impedance sensing interface based on polyethylene glycol (PEG)-modified magnetic nanoparticles (M-PEG) was developed to enable highly selective enrichment and raw detection of CLB by electrochemical impedance spectroscopy. To address the limitations of traditional equivalent circuit models in characterizing complex interfacial behaviors of EIS data, the distribution of relaxation times (DRT) approach was employed to convert frequency-domain impedance data into time-domain relaxation spectra, allowing effective separation of overlapping electrochemical processes and further enhanced feature representation, the one-dimensional DRT sequences were then transformed into two-dimensional images using Gramian Angular Difference Field (GADF) encoding, and a ConformerMultiScaleCNN regression model combining multi-scale convolution with self-attention was proposed to jointly capture local texture patterns and global temporal dependencies. Experimental results indicated that PEG-modified electrodes markedly improved the adsorption capacity and interfacial response toward CLB, demonstrating a well-defined linear response ranging from 1 to 100 μg/L, with a detection limit as low as 0.17 μg/L while the GADF-Conformer model achieved excellent concentration-prediction performance, with an R2 of 0.9952, a root-mean-square error of 1.77 μg/L, and a mean absolute error of 2.33 μg/L. The model also demonstrated strong robustness under label-reduced scenarios and in distinguishing low-concentration targets. Grad-CAM visualization shows that the key attention regions of the model aligned well with the coupled features of double-layer capacitance and charge-transfer processes in the DRT spectra, supporting the interpretability of the proposed approach. Furthermore, tests in artificial sweat and diluted human serum confirm the high accuracy and reproducibility, show great potential for on-site food safety monitoring and anti-doping applications.