Deep Learning-Assisted Rapid Assessment of Food Freshness Using an Anti-interfering Triple-Emission Ratiometric Fluorescent Sensor

Abstract

The assessment of food freshness is of paramount significance for the maintenance of human health. However, the presence of an interfering background signal from food samples often leads to inevitable false negative results, which remains a formidable challenge in the rapid assessment of food freshness. To address this issue, a bioinspired anti-interfering triple-emission ratiometric fluorescent sensor was developed based on a deep learning strategy to enhance the signal-to-noise ratio in complex real sample and to allow for the rapid real-time detection with significantly reduced sample size. It was enriched with tubular foot-like functional groups (–NH2 and –COOH), which showed good linearity between pH 2.5–9.5 with successive fluorescence color change from blue-green to light green, light yellow, orange, and red. Three YOLO deep learning algorithm models were used to construct self-designed smart WeChat applets for high-throughput analysis, and two unique 3D printing toolboxes based on a 96-well plate and cuvette for sample analysis were also designed. The rapid high-throughput classification of a wide range of beverages and real-time monitoring of food freshness based on a hydrogel tag were also validated for reference. Prospectively, deep learning-assisted creation of proportional sensors will be critical to increasing the diversity and high throughput of real-time monitoring.