JPGN Journal Club July 2026: AI in Medicine – From Disease Trajectories to HCC Risk Prediction
Welcome back to ESPGHAN Journal Club! We have another Manager’s Special at the ESPGHAN podcast supermarket – Aisle 2, listeners! Two at once, tag-team interviewing: Prof Dr Carolin Schneider of the Rheinisch-Westfälische Technische Hochschule Aachen, in Aachen (soon to take up a professorship in Dresden), and Dr Jake Mann of Birmingham.
Dr Mann is… well, he’s the bloke that you all know – or so let’s hope! – from faithful attendance at ESPGHAN Journal Club. No need to rabbit on about him. Prof Schneider is a new rocket, or rather perhaps someone riding a new rocket. She comes from Neuss, the garden spot of the Ruhrpott, across the River Rhine from Düsseldorf – and she stayed close to home for her medical training: Lured irresistibly by Printen, a regional type of gingerbread and the emblem of the city of Aachen – Charlemagne’s capital – she headed a few kilometres to the southwest, matriculated at RWTH Aachen University, and took all the prizes. She now heads a laboratory at her alma mater, administratively within the department of gastroenterology. Her Printen-fuelled career-rocket is powered by artificial intelligence (AI), that is to say, by the use of AI to mine large data sets for features that may be of value in population screening.
She and Jake have recommended two articles for listener attention: “Learning the natural history of human disease with generative transformers” and “Machine learning predicts hepatocellular carcinoma risk from routine clinical data: A large population-based multicentric study”, the latter from her team. The former describes use of GPTs – generative pre-trained transformers; a form of AI – to model health trajectories for populations and to identify biomarker combinations of predictive value for morbidity of multiple aetiologies. The latter uses AI as well, again seeking biomarker combinations, but those associated predictively with a particular single disorder – hepatocellular carcinoma.
Now, a recommendation: Seek out these publications (elsewhere on the ESPGHAN website is a précis of this podcast, with bibliographic data). Pore over them. Then come back to the podcast and listen again. Why? To cite Mark Twain: “The statements are interesting, but tough.” They are written in Epidemiology and in Statistics, neither of which is a language that many of us speak. And to plough through these articles and to emerge at the other end feeling that one has understood them… is tough.
Prof Schneider and Jake in the following discussion are excellent guides on a listener’s pilgrimage to the Celestial City of comprehension – yes, the metaphor is from John Bunyan, as were the statements on which Twain commented. But only to a certain extent can their unaided efforts make the crooked straight and the rough places plain. So aid them – read, ponder, and make time to listen to this podcast more than once.
Literature
Clusmann J et al. Machine learning predicts hepatocellular carcinoma risk from routine clinical data: A large population-based multicentric study. Cancer Discov 2026 Mar 26. DOI: 10.1158/2159-8290.CD-25-1323. Online ahead of print. PMID: 41881847
Shmatko A et al. Learning the natural history of human disease with generative transformers. Nature 2025 Nov;647(8088):248–256. DOI: 10.1038/s41586-025-09529-3. Epub 2025 Sep 17. PMID: 40963019. PMCID: 12589094. Erratum: Nature 2025 Nov;647(8091):E8. DOI: 10.1038/s41586-025-09879-y. PMID: 42225015