Extracellular vesicles (EVs) play a crucial role in the occurrence and progression of cancer. The efficient isolation and analysis of EVs for early cancer diagnosis and prognosis have gained significant attention. In this study, for the first time, a rapid and visually detectable method termed freeze-thaw-induced floating patterns of gold nanoparticles (FTFPA) is proposed, which surpasses current state-of-the-art technologies by achieving a 100 fold improvement in the limit of detection of EVs. Notably, it allows for multi-dimensional visualizations of EVs through site-specific oligonucleotide incorporation. This capability empowers FTFPA to accurately identify EVs derived from subtypes of breast cancers with artificial intelligence algorithms. Intriguingly, learning the freezing-thawing-microstructures of EVs with a random forest algorithm is not only able to distinguish their original cell lines (with an accuracy of 95.56%), but also succeed in processing clinical samples (n = 156) to identify EVs by their healthy donors, breast lump and breast cancer subtypes (Luminal A, Triple-negative breast cancer, and Luminal B) with an accuracy of 83.33%. Therefore, this AI-empowered micro-visualization method establishes a rapid and precise point-of-care platform that is applicable to both fundamental research and clinical settings.
Keywords: AuNPs; artificial intelligence; extracellular vesicles; freezing‐thawing.
© 2024 Wiley‐VCH GmbH.