Deep Learning-Enhanced Lateral-Flow Microarray Biosensor for Multiplexed Inflammatory Biomarker Analysis

Abstract

Beyond simply functional integration, achieving synergistic enhancements in multiplex detection, quantitative robustness, and intelligent analysis remains a key challenge for next-generation biosensing platforms in precision diagnostics. Lateral flow biosensors are attractive paper-based platforms because of their simplicity, low cost, and rapid operation; however, conventional line-type formats and visual readouts may limit their multiplex quantitative capability. Here, we present a surface-enhanced Raman scattering-based lateral flow microarray chip (LF-SERS-MAC) platform that integrates spatially encoded microarray patterns with AI-assisted analysis to enable sensitive, specific, and stable detection of three inflammatory proteins on a single paper device. The platform employs AuNPDTNB@antibody nanoprobes that simultaneously provide strong Raman signals and specific recognition of target proteins. The 4 × 4 microarray architecture enables spatially separated detection channels and predefined replicate microdots, supporting standardized SERS sampling and replicate-based signal averaging. More importantly, a 2D transformation-based deep learning model was developed for inflammation classification and concentration prediction, which demonstrates higher accuracy than conventional AI models. This platform can be further scaled to high-throughput paper chips, facilitating its clinical translation as a next-generation multiplexed intelligent biosensing strategy.