Ratiometric fluorescence sensor based on deep learning for rapid and user-friendly detection of tetracycline antibiotics

• Published in: Food chemistry Vol. 450; p. 138961
• Main Authors: Chen, Zhengjie, Li, Zhi, He, Haibin, Liu, Juhua, Deng, Junjie, Jiang, Lin, Liu, Xinghai
• Format: Journal Article
• Language: English
• Published: England Elsevier Ltd 30.08.2024
• Subjects: Anti-Bacterial Agents - analysis; Antibiotic detection; Carbon dots; Colorimetry - instrumentation; Colorimetry - methods; Deep Learning; Fluorescence; Food Contamination - analysis; Intelligent detection; Ratiometric fluorescent hydrogel sensor; Spectrometry, Fluorescence - methods; Tetracycline - analysis; Ratiometric fluorescent hydrogel sensor; Deep learning; Intelligent detection; Carbon dots; Antibiotic detection
• ISSN: 0308-8146; 1873-7072
• DOI: 10.1016/j.foodchem.2024.138961

The detection of tetracycline antibiotics (TCs) in food holds great significance in minimizing their absorption within the human body. Hence, this study aims to develop a rapid, convenient, real-time, and accurate detection method for detecting antibiotics in an authentic market setting. A colorimetric fluorescence sensor was devised for tetracycline detection utilizing PVA aerogels as the substrate. Its operating principle is based on the IFE effect and antenna effect. A detection device is designed to capture fluorescence images while deep learning was employed to aid in the detection process. The sensor exhibits high responsiveness with a mere 60-s requirement for detection and demonstrates substantial color changes(blue to red), achieving 99% accuracy within the range of 10–100 μM with the assistance of deep learning (Resnet18). Real sample simulation tests yielded recovery rates between 95% and 130%. Overall, the proposed strategy proved to be a simple, portable, reliable, and responsive solution for rapid real-time TCs detection in food samples. •Double-emission fluorescent probes were used, and the color variation was obvious.•Hydrophilic hydrogels are used as substrates to improve sensitivity.•More than 1000 sample images enhance the credibility of their deep-learning models.•The deep learning model can detect tetracycline antibiotics with 99% accuracy.