The combination of machine learning and liquid biopsy is rapidly promoting the development of precision cancer diagnosis. Extracellular vesicles and particles (EVPs) are liquid biopsy markers with great diagnostic value due to their unique structure, high stability and strong disease specificity in body fluids. With the rapid development of artificial intelligence (AI), machine learning enables the identification of informative biomarkers from high-dimensional, complex, and large-scale biological data. Thanks to this, EVPs have made rapid progress in the correlation research of tumor diagnosis and prognosis evaluation in recent years. This review focuses on cancer, the area in which liquid biopsy is most extensively studied and clinically needed. This review focuses on cancer as a primary application domain of liquid biopsy and provides a structured overview of machine learning methodologies in this context. We summarize recent advances in feature mining, multi-omics integration, and multi-marker fusion strategies for blood- and EVP-derived data, while also discussing key challenges, including data heterogeneity, model interpretability, and clinical validation.
Yu et al. (Sat,) studied this question.