An Artificial Intelligence System for ROSE of EUS-FNA Sample: a Prospective, Multicenter, Diagnostic Study.

An Artificial Intelligence System for Rapid Onsite Cytologic Pathology Evaluation（ROSE） of Endoscopic Ultrasound-guided Fine-needle Aspiration (EUS-FNA) Sample: a Prospective, Multicenter, Diagnostic Study.

This is an observational study with a prospective, multicenter, disgnostic design. An artificial intelligence system named ROSE-AI system was developed using cytopathological slide images taken by microscope camera or smartphone of pancreas, bile duct, liver and lymph node, collected retrospectively from patients who underwent EUS-FNA and ROSE, and the performance of ROSE-AI system was validated in the datasets collected prospectively.This study aims to assist endoscopists in conducting rapid on-site cytopathology evaluations during EUS-FNA without the presence of cytopathologists. In addition, the diagnostic field was compared between the cytopathologists and ROSE-AI system, endoscopists with or without ROSE-AI system.

Inclusion Criteria:

1. the patient age ≥18 years accepted EUS-FNA+ROSE.
2. agree to participate in the research and be able to sign written informed consent.

Exclusion Criteria:

1. uncorrectable coagulopathy (PTT >50 seconds or INR >1.5) and/or uncorrectable thrombocytopenia (platelet count <50 × 109 /L).
2. patients who were too clinically ill to undergo an EUS examination.
3. lesions that were deemed inaccessible for EUS-guided sampling.
4. unsuccessful EUS-FNA (e.g., failure to obtain an adequate specimen, patient intolerance, intraoperative accidents, etc.).
5. Patients with unqualified ROSE smear.
6. Patients who underwent biopsy during EUS-FNA but did not receive a definitive pathological diagnosis or pathological report.
7. pregnancy.