Development of a novel colorimetric sensor array based on oxidized chitin nanocrystals and deep learning for monitoring beef freshness

Abstract

Food spoilage not only causes food waste, but also leads to serious foodborne illnesses. To address these concerns, a novel colorimetric sensor array (CSA) fabricated from oxidized chitin nanocrystals (O-ChNCs) combined with convolutional neural network (CNN) had been developed for real-time monitoring of beef freshness. The oxidation procedure downsized chitin to the nanoscale with enlarged surface areas and more sites that are reactive. It also converted hydroxyl groups to carboxyl groups, giving chitin nanocrystals high negative charges. These enhanced electrostatic interactions of O-ChNCs with ammonium cations, which was validated with three representative food spoilage gases (methylamine (MA), trimethylamine (TMA), and ammonia (NH3)). The detection limits reached 100, 70, and 70 ppm for MA, TMA, and NH3, respectively. All four CNN architectures succeeded with over 96 % accuracy in discriminating freshness when monitoring real beef samples stored at room temperature for 4 days in real time. The highest accuracy of 99.27 % was achieved by ResNet-50. The overall results indicated that the newly developed O-ChNCs-based CSA coupled with CNN achieved fast and reliable monitoring of beef freshness.