The objective of this project is to pioneer a novel protocol for the adjunctive screening of early-stage esophageal cancer and its precancerous lesions. The anticipated outcomes include simplifying the training process for users, shortening the duration of examinations, and achieving a more precise assessment of the extent of esophageal cancer invasion than what is currently possible with ultrasound technology. This research endeavors to harness the synergy of endoscopic ultrasound (EUS) and Magnifying endoscopy, augmented by the pattern recognition and correlation capabilities of artificial intelligence (AI), to detect early esophageal squamous cell carcinoma and its invasiveness, along with high-grade intraepithelial neoplasia. The overarching goal is to ascertain the potential and significance of this approach in the early detection of esophageal cancer. The project's primary goals are to develop three distinct AI-assisted diagnostic systems: An AI-driven electronic endoscopic diagnosis system designed to autonomously identify lesions. An AI-based EUS diagnostic system capable of automatically delineating the affected areas. A multimodal diagnostic framework that integrates electronic endoscopy with EUS to enhance diagnostic accuracy and efficiency.
The study was executed in two distinct phases. The initial phase, designated as the modeling phase (Phase 1), involved a retrospective analysis of eligible subjects from a consortium of medical institutions, including the First Affiliated Hospital of Naval Medical University, West China Hospital of Sichuan University, Provincial Hospital Affiliated to Shandong First Medical University, the First Affiliated Hospital of Soochow University, the First Affiliated Hospital of Henan University of Science and Technology, and the First Affiliated Hospital of Shihezi University, all selected prior to January 1, 2024. The second phase, known as the real-world evaluation phase (Phase 2), prospectively enrolled consecutive patients who were scheduled to undergo magnometric endoscopy and EUS at the aforementioned hospitals between April 2024 and June 2024.
Study Type
OBSERVATIONAL
Enrollment
450
The acquired magnifying endoscopy and endoscopic ultrasonography images were shared with artificial intelligence for machine learning, diagnostic modeling and optimization. In the real world evaluation phase, the high-risk population of early esophageal cancer who planned to undergo esophageal electronic endoscopy were prospectively enrolled. The artificial intelligence-assisted diagnosis system was used for prediction before surgery, and the postoperative pathological results were used as the gold standard to diagnose by grouping.
Changhai hospital
Shanghai, China
Performance of models to diagnose low-grade intraepithelial neoplasia, high-grade intraepithelial neoplasia, and superficial esophageal squamous carcinoma
Endoscopic Submucosal Dissection (ESD) serving as the gold standard. Computation of sensitivity and specificity involves the use of four fundamental metrics: true positive (TP), true negative (TN), false negative (FN), and false positive (FP). Subsequently, the Area Under the Curve (AUC) is utilized to assess the diagnostic efficacy of the model.
Time frame: 2024.04.01-2024.10.30
This platform is for informational purposes only and does not constitute medical advice. Always consult a qualified healthcare professional.