Hepatocellular carcinoma (HCC) is a leading cause of global cancer-related mortality, with delayed diagnosis adversely affecting patient outcomes. Liquid biopsy techniques using small extracellular vesicles (EVs) offer potential for cancer detection, though current methods are often time-consuming and require complex equipment, limiting clinical utility. Here, we report a metasurface-enhanced EV detection chip (metaEVchip) platform for the dynamic monitoring of HCC-specific EVs, enabling rapid detection and purification. This system provides results within 5 min. The platform integrates a plasmonic metasurface with a Kolmogorov-Arnold network (KAN) to facilitate real-time EV capture, enhancing detection speed while achieving an area under the curve (AUC) of 0.914 for HCC screening. By optimizing the purification process and incorporating complementary detection of alpha-fetoprotein (AFP) and protein induced by vitamin K absence or antagonist II (PIVKAII), the AUC for HCC screening reaches 0.961 in an external validation set. These results effectively differentiate HCC from benign liver diseases (BLD) and early-stage HCC from cirrhosis, addressing limitations of conventional EV detection and demonstrating the potential for rapid cancer screening.
Keywords: cancer screening; deep learning; extracellular vesicles; plasmonic metasurface; rapid detection.