Coupling machine learning with smartphone-assisted sensing platform: sample-to-answer visual detection of amoxicillin using ratiometric fluorescence carbon dots

A fully integrated handheld sensor with a sample-to-answer fashion for target determination is essential for food industry and environmental science. However, the most important issue for existing sensors is still time-consuming and lacks user-friendly on-site monitoring capabilities. To address this issue, we first design a portable artificial intelligence-driven handheld sensor to achieve in-situ monitoring of amoxicillin (AMO). The intelligent platform consists of the high-throughput paper sensor loaded with carbon dots (CDs), a custom-built 3D-printed attachment, and a smartphone with machine learning (ML) function. Note that the introduction of AMO into CDs system triggers a hydrogen bond reaction, resulting in a ratiometric signal variation accompanied by the fluorescence (FL) color changes from yellow to blue. In real-time mode, an open-source ML algorithm is deployed in a self-programming applet from smartphone to directly output AMO concentration of samples, allowing for intelligent on-site analysis. Importantly, the system enables rapid detection of AMO in a wide range of 5 μM to 1000 μM, with an acceptable detection limit of 2.8 μM. Moreover, we verify the integrated system can be used to quantify AMO in real samples. This proposed ML-driven sample-to-answer sensing strategy will inspire the development of intelligent sensors.