The Role of Artificial Intelligence in Endoscopic Diagnosis of Esophagogastric Junctional Adenocarcinoma：A Single Center, Case-control, Diagnostic Study
Trial Snapshot
- Phase
- Not Applicable
- Status
- Not yet recruiting
- Enrollment
- 200
- Primary Endpoint
- Specificity
Study Overview
Brief Summary
This is a single center, case-control, diagnostic study.The aim of this study is to use deep learning methods to retrospectively analyze the imaging data of gastrointestinal endoscopy in Qilu Hospital, and construct an artificial intelligence model based on endoscopic images for detecting and determining the depth of invasion of esophagogastric junctional adenocarcinoma.This study will also compare the established AI model with the diagnostic results of endoscopists to evaluate the clinical auxiliary value of the model for endoscopists.The research includes stages such as data collection and preprocessing, artificial intelligence model development, model testing and evaluation. The gastroscopy image dataset constructed by this research institute mainly includes three modes of endoscopic imaging: white light endoscopy, optical enhancement endoscopy (OE), and narrowband imaging endoscopy (NBI).
Study Design
- Study Type
- Observational
- Observational Model
- Case Control
- Time Perspective
- Retrospective
Eligibility Criteria
- Ages
- 18 Years to 75 Years (Adult, Older Adult)
- Sex
- All
- Accepts Healthy Volunteers
- Yes
Inclusion Criteria
- •This study included endoscopic images of patients aged 18 and above who underwent endoscopic examination or treatment
- •All patients in the case group need to be pathologically confirmed as esophageal gastric junction adenocarcinoma, and a pathologist has conducted a standardized pathological evaluation of the tumor classification of the lesion, including the overall appearance, size, differentiation type, depth of infiltration, presence or absence of lymphatic/vascular invasion, surgical margin status, etc.
- •The endoscopic images of the control group patients need to be confirmed by biopsy pathology or at least two experienced endoscopists (with operating experience>5000 cases) to jointly confirm that they have clear benign manifestations
Exclusion Criteria
- •The patient has a previous history of endoscopic treatment or surgery for the esophageal gastric junction.
- •Necessary clinical information cannot be provided during the research process (patient age, gender, lesion characteristics, endoscopic manifestations, endoscopic images, etc.)
- •Low quality endoscopic images, such as those severely affected by bleeding, aperture, blurring, defocusing, artifacts, or excessive mucus after biopsy.
Outcomes
Primary Outcomes
Specificity
Time Frame: 36 months
The researchers calculated the specificity of the established AI model and compared it with endoscopists of different levels.
Positive predictive value
Time Frame: 36 months
The researchers calculated the positive predictive value of the established AI model and compared it with endoscopists of different levels.
Sensitivity
Time Frame: 36 months
The researchers calculated the sensitivity of the established AI model and compared it with endoscopists of different levels.
Accuracy
Time Frame: 36 months
The researchers calculated accuracy positive predictive value of the established AI model and compared it with endoscopists of different levels.
Negative predictive value
Time Frame: 36 months
The researchers calculated the negative predictive value of the established AI model and compared it with endoscopists of different levels.
Secondary Outcomes
No secondary outcomes reported