SIGINIFCANTLY improved lesion detection and reduced interpretation times during capsule endoscopy procedures thanks to an open-source artificial intelligence (AI) model in multicentre study.
Researchers evaluated SEE-AI, a pretrained AI model
designed to assist clinicians reviewing small bowel capsule endoscopy videos against conventional review methods. 249 PillCam SB3 capsule endoscopy examinations performed across six hospitals between 2007–2022 were retrospectively analysed using a two-reader crossover design.
AI Assisted Review Enhanced Lesion Detection and Cuts Capsule Reading Time
SEE-AI generated annotated videos with highlighted lesion areas across eight lesion categories. Across 1,550 confirmed lesions, AI-assisted capsule endoscopy achieved significantly higher sensitivity than conventional reading.
Per-lesion sensitivity increased from 86.4% with standard interpretation to 98.8% using SEE-AI-assisted review. On a per-patient basis, sensitivity improved from 80.3% to 99.1%. Researchers also reported a significant reduction in mean reading time, decreasing from 17.9 minutes to 13.7 minutes.
Benefits were particularly notable among patients with suspected small-bowel bleeding (SSBB). In this subgroup, sensitivity for clinically relevant bleeding lesions improved from 82.8% to 98.2% on a per-lesion basis and from 73.5% to 98.6% per patient. Reading times were also reduced from 18.0 minutes to 14.1 minutes.
Open-Source AI Model Could Support Future Standard of Care
SEE-AI remained fully open-source, supporting transparency and reproducibility in AI-assisted gastrointestinal imaging. The authors suggested that integrating AI into capsule endoscopy workflows may help reduce clinician workload while improving diagnostic accuracy.
Researchers suggested that further prospective investigation will be needed to determine how AI-assisted capsule endoscopy performs in routine clinical practice across broader patient populations.
The researchers concluded that SEE-AI demonstrated strong potential as a practical support tool and may contribute to establishing AI-assisted capsule endoscopy as a future standard of care for small bowel imaging.
Reference:
Miyazono S et al. Improved efficiency and lesion detection in small bowel capsule endoscopy using the open-source artificial intelligence model SEE-AI. DEN Open. 2026;7:e70346. DOI:10.1002/deo2.70346.
Featured Image: wedmoments.stock on Adobe Stock
