AI-assisted label-free single-particle analysis of milk-derived extracellular vesicles enabled by nanotweezers

Milk-derived extracellular vesicles (mEVs) are promising drug carriers because they can survive the digestive tract, be engineered to avoid immune rejection, and deliver therapeutic cargo to the body. Fully realizing this potential requires precise, single-particle measurements of size, composition, and purity, yet current analytical tools are often slow, require chemical labels. Here we present an electrohydrodynamic nanotweezer platform that rapidly traps thousands of mEVs in parallel, enables label-free interferometric scattering imaging, and uses artificial intelligence for automated analysis. The device employs a thin gold film patterned with a 15 µm hole array that generates radially-converging AC-driven fluid flows to hold vesicles in place in parallel within seconds. By briefly releasing them, we track their Brownian motion to calculate their size from diffusion behavior and estimate refractive index from their interferometric contrast signals to assess sample purity and heterogeneity. This fast, non-perturbative workflow combines scalable trapping, real-time label-free imaging, and automated analysis, offering a powerful approach to studying mEVs and accelerate their use as therapeutic delivery vehicles.