AI in Upper GI: Investment or Expense?
Having asked what to scope and how, the day turns to what to buy — and this is where the meeting's both-directions rule gets tested in public. Not "resist the purchase", but: what would make this purchase right?
Running order (indicative)
- Head
- Dr John Andersonanchors the studio throughout
- Talk 1
- Prof Alanna EbigboUpper-GI AI in 2026: What the Trials Show — and What They Don't
- Talk 2
- Prof Raf BisschopsBeyond the Trial Data: A Buyer's Framework for AI, With Kill Criteria
- Live case
- Prof David TateDr Philippe LeclercqSurveillance / diagnostic gastroscopy run with AI (CADe + quality) live on screen
- Discussion
- Allremaining session time
The session, as written
Lecture 1 — Upper-GI AI in 2026: what the trials show, and what they don't. (Ebigbo). Barrett's and gastric detection are promising but behind colonic CADe in deployment and outcome data. Two honest riders: the value of AI sits with the generalist on a run-of-the-mill list, not the expert — and it must never substitute for training or hand trainees a false sense of security.
Lecture 2 — Beyond the trial data: a buyer's framework, with kill criteria. (Bisschops). Trials say one thing, vendors another, your budget a third: most units cannot buy this today, so what would change that? Licence and per-procedure cost, integration, audit endpoints, return on investment — and explicit kill criteria if nothing measurable has moved by twelve months.
Live case — AI-assisted gastroscopy. Detection and quality tools running live: what the AI flags, what it misses, and whether it changed a single decision during the procedure.
