Artificial intelligence-assisted monitoring of arteriovenous fistula function in hemodialysis: Opportunities and challenges

This is the authors' abstract. We don't add key points for this paper.

Reliable vascular access is essential for maintenance hemodialysis. In appropriate patients, arteriovenous fistulas (AVFs) provide durable access, but remain vulnerable to maturation failure, stenosis, thrombosis, and recurrent dysfunction. Routine clinical monitoring and clinically indicated instrument-based assessments are valuable but intermittent, creating an opportunity for repeated, low-burden physiological monitoring between encounters. Artificial intelligence (AI)-assisted systems analyze acoustic, optical, mechanical, thermal, impedance, and multimodal signals that represent physiological proxies rather than stenosis itself. This review examines their use in dialysis-unit screening and triage, maturation assessment, post-angioplasty follow-up, and selected longitudinal or home monitoring within a pathway from reproducible measurement to AI-supported interpretation and clinical reassessment. Currently, acoustic sensing provides the most developed near-term pathway for clinical evaluation, because AVF-specific human evidence and workflow compatibility extend across prospective cohorts, multicenter pooled development, and longitudinal studies. Current evidence is constrained by heterogeneous tasks and analysis units, predominantly internal validation, and insufficient evidence regarding calibration, signal failure, false alert burden, workflow utility, and clinical impact. Progress requires standardized acquisition, patient- or access-independent validation, genuine external evaluation, prospective silent deployment, and clinical impact trials. Routine adoption should depend on evidence that the complete sensing-to-action pathway improves vascular access care or patient-important outcomes without disproportionate workload or harm.