Generative Chromoendoscopy for Detecting Early Gastric Neoplasms: Endoscopy | July 2026
Introduction:
Conventional chromoendoscopy enhances the detection of early gastric neoplasms but requires dye spraying, adding time and procedural complexity. This proof-of-concept multicenter study evaluated Generative Chromoendoscopy, an artificial intelligence–based technique that uses deep learning to convert standard white-light images into virtual chromoendoscopy images without applying dye.
Why was this study needed?
- Early gastric cancer can be subtle and difficult to detect with white-light endoscopy.
- Conventional indigo carmine chromoendoscopy requires dye application and prolongs procedures.
- Artificial intelligence may generate enhanced images without additional equipment or dye.
- Clinical feasibility of AI-generated chromoendoscopy had not previously been validated.
- Improved lesion visibility could facilitate earlier diagnosis and treatment.
Results:
- Generative chromoendoscopy significantly improved the visibility of early gastric neoplasms compared with conventional white-light imaging and achieved visibility comparable to real indigo carmine chromoendoscopy.
- Differentiated-type gastric neoplasms, active Helicobacter pylori infection, and expert endoscopist interpretation were associated with the greatest improvement in lesion visibility.
- The study demonstrated that AI-generated chromoendoscopy was feasible during live endoscopic procedures, supporting its potential integration into routine endoscopic practice.
Clinical Impact:
This proof-of-concept study introduces generative chromoendoscopy as a novel application of generative artificial intelligence in gastrointestinal endoscopy. If validated in larger studies, it could provide real-time image enhancement without dye spraying, simplifying endoscopic workflows while improving early gastric cancer detection.
Bottom Line:
Generative chromoendoscopy uses artificial intelligence to transform standard white-light endoscopic images into virtual chromoendoscopy, improving the visibility of early gastric neoplasms without the need for dye application. This technology represents a promising step toward AI-assisted image enhancement in diagnostic endoscopy, although larger validation studies are needed before widespread clinical adoption.
