Artificial Intelligence (AI) in Gastrointestinal Capsule Endoscopy: Hardware, Algorithmic Applications, and Clinical Integration in healthcare.
Gastrointestinal (GI) capsule endoscopy has evolved from a passive ingestible imaging device into an increasingly automated diagnostic platform supported by artificial intelligence (AI). By generating tens to hundreds of thousands of images per examination, capsule procedures create a substantial interpretation burden and opportunity for AI-assisted image triage, lesion detection, and workflow optimization. This narrative review evaluates the enabling hardware architecture of contemporary GI capsule endoscopy systems; summarize AI applications in small-bowel and colon capsule endoscopy; and identify the clinical engineering, regulatory, validation, and implementation issues relevant to safe adoption. This narrative technology review synthesizes peer-reviewed literature, selected technical sources, manufacturer documentation, and regulatory materials concerning capsule endoscopy hardware, AI-assisted image analysis, clinical performance, and deployment considerations.
This review also explored the mathematical and machine-learning algorithms utilized in capsule technology. Contemporary capsule endoscopes integrate miniature optical systems, complementary metal-oxide-semiconductor image sensors, light-emitting diode illumination, battery power, radio-frequency telemetry, onboard control electronics, and biocompatible packaging within a constrained ingestible form factor. Power, size, thermal, transmission, and computing limitations generally require AI analysis to occur after image acquisition on a workstation or cloud-connected platform rather than within the capsule. Deep-learning methods—predominantly convolutional neural networks—have been applied to frame triage, bleeding and vascular-lesion detection, ulcer and erosion identification, protruding-lesion classification, inflammatory bowel disease assessment, polyp detection, bowel-preparation scoring, and automated reporting support. Small-bowel applications are comparatively mature, whereas colon capsule AI remains less developed for lesion localization, cross-frame matching, size estimation, and patient-level clinical validation. AI-enabled capsule endoscopy should be assessed as an integrated clinical system rather than as an image-classification algorithm alone. Successful deployment requires verification of intended use and regulatory status, device–software compatibility testing, cybersecurity and data-governance controls, reader training, workflow design that mitigates automation bias, software version control, and post-market performance monitoring.
KEYWORDS: Artificial intelligence (AI), Capsule endoscopy (CE), Wireless capsule endoscopy (WCE), Small-bowel capsule endoscopy (SBCE), Colon capsule endoscopy (CCE), Pan-enteric capsule endoscopy, Convolutional neural network (CNN), Computer-aided detection (CADe), Deep learning (DL).
