For adults navigating pancreatic disease, liver conditions, or gastrointestinal cancers, the tools physicians use to detect and characterize disease are undergoing a quiet revolution. Endoscopic ultrasound — long considered a specialist's imaging workhorse — is rapidly transforming into a multi-modal diagnostic and therapeutic platform with implications for earlier cancer detection and more precise tissue sampling.
This review in Gastroenterology maps the frontier of endoscopic ultrasound (EUS) across several high-stakes clinical domains. In pancreatic cancer surveillance, secretin-stimulated collection of duodenal pancreatic juice enables liquid biopsy analysis — specifically methylated DNA markers and extracellular-vesicle microRNAs — that can work alongside the conventional CA 19-9 blood marker to potentially detect malignancy earlier than standard imaging. For chronic pancreatitis, EUS remains the preferred tool when cross-sectional scans are inconclusive, though discordant functional testing results temper its standalone diagnostic certainty. In what is being called endohepatology, EUS-guided portal pressure gradient measurement and same-session liver biopsy now offer accurate hepatic assessment with strong safety data. Tissue acquisition has been refined through end-cutting fine-needle biopsy needles, fanning techniques, and wet-suction approaches that reduce needle passes while maintaining molecular sample quality. Artificial intelligence is being integrated across EUS workflows — from lesion classification and cancer staging to automated photodocumentation.
These advances collectively represent a meaningful shift rather than an incremental refinement. Liquid biopsy via pancreatic juice collection is particularly noteworthy: pancreatic cancer carries a five-year survival rate under 15%, largely because most cases are caught late. Any tool that improves early detection in high-risk populations — those with new-onset diabetes, family history, or pancreatic cysts — addresses a genuine unmet need. The hepatology applications also matter, since non-invasive portal pressure measurement has historically required invasive hepatic vein catheterization. Key limitations include the procedural expertise required, limited multicenter validation for AI models, and the fact that most liquid biopsy data in this context remain exploratory rather than practice-defining. This is a field moving fast, but clinical adoption will depend on prospective outcome data.