A Machine Learning-assisted Smartphone-based Method for Rapid Quantification of Total Flavonoids by Quantum Dots

This study presents a rapid and instrument-free method for the quantification of total flavonoids, utilizing smartphone imaging and lightweight machine learning techniques. The methodology involves capturing fluorescence images of zinc-based quantum dots that are quenched by varying concentrations of quercetin, a representative flavonoid, using a smartphone camera. A pre-trained convolutional neural network (CNN) model was developed, which can be fine-tuned with a minimal dataset to achieve accurate quantification of total flavonoids in specific plant samples. Using Euphorbia humifusa as a model plant, the method demonstrated a linear detection range of 5–500 μg/mL, with quantification errors ranging from 0.1 to 10 μg/mL. Compared to conventional methods, the detection limit was improved by a factor of 1.6, and the upper detection limit was also extended by a factor of 1.6. This approach provides simplicity, cost-effectiveness, and independence from specialized instruments, thereby offering a promising solution for on-site rapid detection of total flavonoids in plants.