Artificial intelligence and extracellular vesicles in oncology: towards tumor diagnosis, prediction, and therapy

Extracellular vesicles (EVs) have emerged as promising tools for early cancer detection, therapeutic monitoring, and drug delivery in oncology. Artificial intelligence (AI), particularly machine learning and deep learning, offers new analytical tools and computational approaches for EV research. This review summarizes recent advances in the application of AI to EV isolation, characterization, diagnosis, and drug delivery, with particular emphasis on its potential to enhance tumor detection sensitivity, diagnostic accuracy, and the rational design of delivery platforms. Special attention is given to the roles and recent applications of AI models in integrating multimodal features, characterizing EV heterogeneity, supporting diagnostic classification, and modeling in vivo behavior. Moreover, we examine the integration of AI with EV-based microfluidic isolation, surface-enhanced Raman spectroscopy (SERS), fluorescence imaging, and multiomics analysis. Among these areas, AI-assisted EV diagnostic applications are comparatively closer to clinical translation, with several studies incorporating patient-derived samples and AI-assisted diagnostic platforms, whereas AI-guided therapeutic EV design strategies remain largely exploratory. With the continued accumulation of multicenter, cross-platform EV datasets, improvements in algorithmic robustness, and closer integration of computational and experimental workflows, AI may support further clinical evaluation of EV-based diagnostics and the systematic optimization of therapeutic EV platforms.