AI-Powered Surface Light Microscopy for In Situ Profiling of Cell Adhesions and Membrane Protein Binding Kinetics.

Ligand binding to membrane proteins initiates signaling and enables therapeutic intervention, positioning these receptors as central targets in biology and drug development. Although surface plasmon resonance (SPR) imaging allows label-free, in situ observation of molecular interactions at the membrane, quantitative analysis is often compromised by interference from background cellular activity. To overcome this limitation, we introduce an artificial intelligence (AI)-powered surface light microscopy platform that integrates a waveguide sensor chip for dual-channel imaging and an AI-driven recognition module. The waveguide generates an evanescent field for surface light scattering imaging as SPR, while simultaneously supporting fluorescence imaging without quenching to spatially profile membrane-associated adhesions. Using dual-channel data, we train an AI model to automatically recognize membrane-associated adhesions in scattering images, achieving an area under the curve of ∼0.85. Once trained, the model operates autonomously on scattering sequences, eliminating the need for further fluorescence input. By selectively analyzing AI-identified regions, we isolate real-time ligand-binding signals from nonspecific background, enabling robust in situ binding kinetic analysis at the single live cell level. This work addresses a fundamental limitation of label-free biosensing by decoupling the sensitivity of physical measurement from the specificity of biological interpretation, transforming these techniques from qualitative, context-sensitive tools into quantitative platforms capable of molecularly precise analysis directly on live cells.