Open-surface digital ELISA enabled by magnetic trapping and deep learning for rapid and highly sensitive detection of African swine fever.

Abstract

African swine fever (ASF), a highly fatal and contagious animal disease caused by ASF virus, leads to substantial economic losses. Early detection is critical, as the viral load is initially low, demanding highly sensitive detection of the viral p30 antigen. Although the existing digital ELISA enables sensitive detection of p30, its performance is constrained by inefficient microcavity utilization (resulting in Poisson noise-limited sensitivity) and low image recognition accuracy. This study presents an open-surface digital ELISA (OS-dELISA) platform that integrates an open-space magnetic bead (MB) array with artificial intelligence. The platform uses magnetic trapping to form the MB array, which acts as an open-space microcavity. Combined with reciprocating-flow microfluidics and in-situ tyramine signal amplification, this design eliminates the additional step of encapsulating immunocomplexes into enclosed microcavities, enhances microcavity utilization and antigen capture efficiency, and improves sensitivity. An improved mask region-based convolutional neural network enables intelligent and accurate image recognition for large-scale datasets. OS-dELISA achieves highly sensitive p30 detection within 17 min, with a detection limit of 21 fg/mL. Serum samples validation showed 100% sensitivity and 95% specificity. By combining microfluidics and artificial intelligence, OS-dELISA provides a powerful tool for intelligent animal disease diagnostics.