A newly developed artificial intelligence (AI) system may help clinicians distinguish benign from malignant gastric ulcers during endoscopy with high accuracy and near-instant analysis speeds, according to a multicentre study.
AI Meets Real-Time Endoscopy
Researchers developed a deep learning model using endoscopic images and videos collected from four hospitals in China. The system was designed to analyse ulcers dynamically during procedures, potentially assisting clinicians in identifying suspicious lesions more accurately.
The model was trained on more than 27,000 images, including benign ulcers, malignant ulcers, and normal gastric mucosa.
High Accuracy for Cancer Detection
On validation testing, the AI system achieved an overall precision and recall of 91%, with specificity reaching 95%.
For malignant gastric ulcers specifically, the model demonstrated particularly strong performance, achieving 90% precision, 91% recall, and 99% specificity.
These findings suggest the tool could reliably identify malignant lesions while limiting false-positive classifications.
Designed for Clinical Workflow
A major advantage of the system is its speed. The AI processed images at 113 frames per second, with a latency of less than 9 milliseconds per frame, allowing true real-time analysis during endoscopy.
This could make the technology practical for integration into routine clinical workflows without slowing procedures.
Supporting Earlier Gastric Cancer Diagnosis
Differentiating malignant gastric ulcers from benign lesions can sometimes be challenging during endoscopy, particularly in subtle or early-stage disease. Delayed or missed diagnoses may impact patient outcomes.
By providing real-time image interpretation, AI tools like this may help clinicians target biopsies more effectively and improve diagnostic consistency between operators.
The Future of AI-Assisted Endoscopy
The researchers suggest the model could support more standardised diagnostic quality across centres while reducing variability in interpretation.
Although prospective real-world studies are still needed, the findings add to growing evidence that AI-assisted endoscopy may play an increasingly important role in earlier cancer detection and precision gastroenterology.
Reference
Tan Y et al. Development and validation of a real-time AI model for differentiating benign and malignant gastric ulcers : a multicenter retrospective study. BMC Gastroenterol.2026;DOI: 10.1186/s12876-026-04848-9
