Endoscopic Ultrasound-guided Fine-needle Aspiration of Solid Pancreatic Lesions With Rapid Staining of Cytological Smears Followed by Whole Slide Scanning and Artificial Intelligence Diagnosis: A Prospective, Multicenter Study.

Condition(s) studied

Investigational drug(s) / intervention(s)

ROSE-AI diagnostic system: All samples were obtained due to the necessity for disease treatment and in accordance with routine clinical workflows. After the pathological diagnoses were confirmed by the pathology departments of the hospitals affiliated with the respective endoscopic centers, the eligible pancreatic puncture Diff-Quik stained smears were borrowed and transferred to Ruijin Hospital Affiliated to School of Medicine, Shanghai Jiao Tong University. There, the self-developed "Zhiying Shunxi" system was used to capture corresponding traditional light microscope RGB images. After the imaging was completed, all specimens were returned to the endoscopic centers from which they originated. Using the RGB images as input, an artificial intelligence algorithm was developed to assist in differentiating solid pancreatic lesions.

Study summary

The objective of this observational study is to investigate whether the self-developed whole slide scanning and artificial intelligence diagnostic system for pancreatic solid lesion puncture cytopathology (hereinafter referred to as the "Zhiying Shunxi" ROSE-AI diagnostic system) can promptly and accurately diagnose solid pancreatic lesions (SPLs). The main question it aims to answer is:

By utilizing optical imaging technology to capture RGB images of Diff-Quik stained smears from pancreatic punctures, can the development of artificial intelligence algorithms assist in differentiating solid pancreatic space-occupying diseases (such as pancreatic ductal adenocarcinoma, pancreatic neuroendocrine tumors, and non-neoplastic benign lesions)?

Researchers will compare the diagnoses of SPLs made by the ROSE-AI system with the actual pathological diagnoses of the SPLs themselves to determine whether the ROSE-AI system can effectively diagnose SPLs.

Eligibility

Primary outcome measure(s)

• Accuracy — through study completion, an average of 2 years
  Accuracy = (TP + TN) / (TP + FP + FN + TN)

Trial sites (2)