BPG is committed to discovery and dissemination of knowledge
Minireviews
Copyright: ©Author(s) 2026.
Artif Intell Gastroenterol. Nov 8, 2026; 7(3): 124278
Published online Nov 8, 2026. doi: 10.35712/aig.124278
Table 3 Artificial intelligence for diagnostic characterisation and prognosis
Ref.
Design, region
Population, aetiology and evidence category (n)
Endpoint
Approach
Validation
Performance
Calibration and explainability
Key limitation
Wei et al[33]Retrospective, ChinaResected HCC; recurrence-related (434)Early recurrence after treatmentMRI radiomics with 3D U-net automated segmentationInternal validation cohortC-index 0.706; exceeded BCLC and CNLC stagingNot reportedPost-diagnosis endpoint; single-country
Ji et al[34]Retrospective, multi-institutional, ChinaResected HCC; recurrence-relatedPost-resection recurrenceMachine learning on contrast-enhanced CT radiomicsMulti-institutionalPredicted recurrence across institutionsNot reportedRetrospective; feature instability across scanners
Zhao et al[35]Retrospective, ChinaHCC after curative resection; recurrence-relatedEarly recurrenceDeep-learning radiomics on contrast-enhanced MRIInternalHybrid models outperformed single-stream inputsNot reportedSingle-centre; no external validation
Chaudhary et al[36]Retrospective, public cohortsLiver cancer; prognosticSurvival and molecular subgroupDeep-learning multi-omics integrationExternal replication in independent cohortsReproducible prognostic molecular subgroupsNot reportedOmics assays costly and unavailable in high-burden settings
Kucukakcali et al[37]Case-control, open-access gene expression dataHBV-related HCC vs HBV infection alone; HBV-specific (17 vs 36)Discrimination of HBV-HCC from HBV aloneExtreme gradient boosting with 10-fold cross-validationInternal cross-validation18 genes selected; candidate biomarkers nominatedNot reportedVery small sample; reported accuracy not an estimate of real-world performance


Write to the Help Desk