Colorimetric microneedle sensor using deep learning algorithm for meat freshness monitoring

Currently, the developed testing methods determining meat freshness are time-consuming, inconvenient, or have high specialty requirements. Herein, we proposed a colorimetric microneedle sensor (CMS) using a deep learning algorithm for visualized meat freshness monitoring. The CMS was obtained by molding edible hydrogels containing pH-responsive anthocyanins, which change colors because of the structure change of anthocyanins in response to pH. When attached to meat, the CMS was capable of penetrating the meat and extracting tissue fluids by capillary action. With meat spoilage, the pH of the tissue fluid gradually rose, leading to a change in CMS from pink to purple and finally to dark blue. Thus, according to variations of CMS colors, in situ and visualized detection of meat freshness was achieved. Further, a deep learning algorithm was applied to integrate with CMS to form a smartphone application (App), allowing for more convenient and accurate freshness detection. Images of CMS attached to the meat with different freshness were collected to form a training source as the input of the convolutional neural network (CNN). Through convolving CMS color features, the meat freshness classified as "fresh", "less fresh", and "spoiled" was finally outputted. With the incorporation of CNN, the App enabled users to identify the freshness of meat from stored photos or real-time images of CMS-labeled meats in a fast, accurate, portable, and universal way. This visualized detection strategy of CMS combined with an algorithm-integrated App has a promising potential for wide applications such as food safety, health monitoring, and environmental protection.