Cure8

Why This Matters

AI tools are being tested to speed capsule endoscopy reading and help triage cases, which matters to people with IBD because capsule endoscopy is commonly used to detect small-bowel ulcers and inflammation.

The study found TOP100 missed some lesions compared with expert readers but had high specificity, so it may help prioritize cases in busy centers.

Who Should Pay Attention

Clinicians who order or read capsule endoscopy, radiologists/endoscopists, IBD patients undergoing small-bowel capsule endoscopy, and researchers developing AI interpretation tools.

Study Snapshot

Story typeResearch paper
Evidence typeResearch paper
Source depthJournal abstract

What To Know

This study compared an AI-assisted capsule endoscopy reading tool (TOP100) to expert human readers, using retrospective data from >1300 PillCam SB3 examinations at a tertiary center (Jan 2023–Aug 2025).

Human readers had higher overall diagnostic yield; TOP100 showed moderate sensitivity but high specificity for key findings (for example, active bleeding sensitivity 64%/specificity 99%; ulcers sensitivity 55.3%/specificity 95.9%). In the subgroup with IBD, ulcer detection by TOP100 was similar to humans and erosion detection improved modestly.

How this might affect patients SR (expert human reading) remains the reference standard; TOP100 could help prioritize cases and highlight obvious lesions in high-volume services but will miss some findings compared with full human review.

This study is an abstract-based summary of a journal article; it reports diagnostic performance metrics rather than patient-level outcomes. Practical note If your care involves capsule endoscopy, ask how AI tools are used at your center—whether they’re used as a first-pass triage, a second reader, or not used at all.

A negative AI-only read does not replace expert review.

Keep In Mind

Structured content depth: abstract. This classification and note are grounded in the article abstract/extracted text. The study is retrospective and compares TOP100 to expert human readers; it reports sensitivity/specificity for lesion types but does not change clinical recommendations. TOP100 may be useful as an adjunct for triage but does not replace standard human reading.

Source Details

Review the original publication for the complete reporting, methods, and context.

Read Original Source
Research paper Evidence type derived from source or registry metadata.
PublicationDiagnostics (Basel, Switzerland)
AuthorsDevanand A, Caranfil C, Moore M +8 more
Study typeJournal article
Indexed viaEurope PMC
Source typeResearch paper
PublishedJun 27, 2026, 12:00 AM
Content availableJournal abstract

This Cure8 brief is based on source text from the linked article. Cure8 is informational only and is not a substitute for professional medical advice, diagnosis, or treatment.

Related Reading

Browse latest news →