Work

AI-integrated SPRi with adaptive microwell-array recognition for high-throughput parallel readout

Year: 2026

Type: article

Abstract: Multiplexed biosensing is vital for modern clinical diagnostics, where Surface Plasmon Resonance Imaging (SPRi) is emerging as a powerful technique for real-time, label-free multi-analyte detection. Conventional SPRi systems however remain limited by optical artifacts, manual data processing, and costly, inflexible hardware. To overcome these challenges, an AI-integrated SPRi platform is presented in this work that combines microarray patterns spanning 200-1000 µm on a single chip, enabling detection of biomarkers across a broad refractive index ranges without hardware modification. The platform transitions from traditional angular-scan SPR to a speckle-free imaging configuration, using microfabricated chips with ~5% variation at wafer-scale for multiplexed sensing. A computational pipeline featuring a fine-tuned Segment Anything Model (SAM2) is applied for the first time in a biosensing context and achieves 83.9% detection recall, representing a 32% improvement over the base model. A novel, two-step algorithm designed to correct optical geometric aberrations restores the microarray patterns to their actual shape, and is fully integrated into the analysis pipeline, which also handles illumination correction, quantitative extraction, and post-processing. The AI-driven pipeline reduces analysis time from hours to minutes and decreases inter-region variability by >65%, ensuring standardized sensitivity across 156 sensing sites while maintaining calibration linearity (R² > 0.94). Together, the integrated hardware and computational framework deliver an accessible, high-throughput, and reproducible SPRi system that unites optical precision with algorithmic intelligence. This work establishes a scalable and standardized foundation for advancing SPRi-based multiplexed biosensing toward practical diagnostic applications.

Source: Sensors and Actuators Reports

DOI: 10.1016/j.snr.2026.100538