Machine learning-assisted NaEuF₄ nanoparticle-enhanced time-resolved fluorescence immunoassay for ultrasensitive and...

An ultra-sensitive time-resolved fluorescence immunoassay (U-TRFIA) based on NaEuF₄ nanoprobes was developed for the detection of aflatoxin B₁ (AFB₁) in grain samples, and machine learning was further integrated to assist in signal interpretation. The synthesized NaEuF₄ nanoprobes, characterized by high Eu³⁺ loading and stability, were used as signal-amplifying probes to label the secondary antibody. Based on an indirect competitive assay, this method achieved an extremely low limit of detection (LOD) of 20 pg/mL, with an IC₅₀ of 34.8 pg/mL and a detection range of 20-500 pg/mL. Compared to conventional TRFIA, this platform offers a fivefold increase in detection sensitivity. Furthermore, the fluorescence images were converted into an RGB feature dataset and analyzed using a machine learning model, establishing a nonlinear mapping relationship between the optical response and AFB₁ concentration. The applicability of the proposed method was validated using grain samples, with predicted concentrations showing good agreement with actual concentrations. Overall, the combination of NaEuF₄-based U-TRFIA with machine learning provides a sensitive, visual sensing platform for AFB₁ detection in food safety applications.