Langue and Imaging-integrated Foundation Model for Gastric Cancer Detection and Staging Via Contrast-Enhanced CT: a Multicenter Study
Trial Snapshot
- Phase
- Not Applicable
- Status
- Recruiting
- Enrollment
- 8,000
- Locations
- 1
- Primary Endpoint
- Diagnostic performance of the AI model for staging
Study Overview
Brief Summary
Accurate preoperative assessment of gastric cancer stage guides eligibility for endoscopic resection, extent of gastrectomy and lymphadenectomy, selection for neoadjuvant therapy, and use of staging laparoscopy. Contrast-enhanced CT (CECT) is guideline-endorsed for initial staging, yet performance varies across institutions and readers. This study will evaluate an artificial-intelligence (AI) system that analyzes routine CECT to detect gastric cancer and assign four-class T stage (T1-T4) and N stage (N0-N3) .
Detailed Description
Adults with confirmed gastric cancer undergoing pre-treatment CECT will be enrolled. The AI analysis will be applied to clinically acquired images. Radiologist interpretations with and without AI support will be collected in a prespecified reader study. The reference standard will include surgical pathology, supplemented by clinical follow-up when applicable. The primary outcome is detection performance, diagnostic performance of the AI for four-class staging (e.g., accuracy and area under the receiver operating characteristic curve). Secondary outcomes include the effect of AI assistance on reader accuracy and interpretation time, inter-reader agreement, and cross-site reproducibility.
Study Design
- Study Type
- Observational
- Observational Model
- Case Only
- Time Perspective
- Retrospective
Eligibility Criteria
- Ages
- 18 Years to 85 Years (Adult, Older Adult)
- Sex
- All
- Accepts Healthy Volunteers
- No
Inclusion Criteria
- •pathologically confirmed gastric cancer;
- •preoperative contrast-enhanced CT performed;
- •no evidence of distant metastasis on baseline staging;
- •curative-intent management with complete postoperative histopathology.
Exclusion Criteria
- •prior treatment before surgery;
- •non-diagnostic or poor-quality CT precluding evaluation.
Outcomes
Primary Outcomes
Diagnostic performance of the AI model for staging
Time Frame: 3 years
The primary outcome is the diagnostic accuracy of the AI system for four-class T staging (T1-T4) and N staging (N0-3) based on contrast-enhanced CT. The AI performance will be assessed using accuracy, area under the receiver operating characteristic curve (AUC), and micro-AUC for internal and external cohorts.
Secondary Outcomes
- Reader Accuracy with AI Support(3 years)
- Survival time(3 years)