Artificial Intelligence as a Surrogate for Inspection Time to Assess Completeness in Esophagogastroduodenoscopy: A Prospective, Randomized, Noninferiority Study
Bibliographic record
Abstract
INTRODUCTION: The completeness of esophagogastroduodenoscopy (EGD) is a prerequisite for detecting lesions. This study aims to explore whether the quality of complete examinations assisted by artificial intelligence (AI) would be comparable with those conducted within the guideline-recommended inspection time. METHODS: Patients referred for diagnostic, screening, or surveillance EGD were enrolled at Renmin Hospital of Wuhan University. Patients were randomly assigned to 2 groups in a 1:1 ratio. In the AI-assisted group, endoscopists completed observation of the entire upper gastrointestinal tract with AI assistance. In the control group, endoscopists were instructed to spend no less than 7 minutes on each procedure. The primary outcome was the detection rate of neoplastic lesions. Noninferiority was confirmed when the lower bound of the 95% confidence interval (CI) was greater than the margin of -1.5%. RESULTS: A total of 1,723 patients were prospectively enrolled between July 3, 2023, and April 7, 2024. Seven hundred ninety-six and 763 patients in the AI-assisted and control groups were included in the final analysis, respectively. The detection rates of neoplastic lesions in the AI-assisted and control groups were 3.14% and 2.36%, respectively, resulting in an absolute proportion difference of 0.78% (95% CI -0.58% to 2.14%; odds ratio 1.342 [95% CI 0.726-2.480]). The median inspection time was reduced by 1.5 minutes in the AI-assisted group (6.18 [2.87] vs 7.70 [1.90], P < 0.001). DISCUSSION: Inspection time of complete EGD can be significantly shortened by AI without compromising its quality. These findings provide crucial evidence to support that AI-assisted procedural completeness serves as an objective and effective quality indicator for EGD.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.017 | 0.018 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.003 | 0.003 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.003 | 0.003 |
| Insufficient payload (model declined to judge) | 0.002 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".