Edinburgh 2026: Gastroenterology and Hepatology

• Personalized medicine is broader than genetics: it combines phenotype, biomarkers, imaging, molecular data, environment and treatment history.

• In GI oncology, molecular profiling already determines therapy for selected patients through biomarkers such as MSI/MMR and actionable genomic alterations.

• Pharmacogenomics has established clinical applications—for example, identifying patients at increased risk of thiopurine toxicity.

• In IBD, the next challenge is predicting response before exposing patients to months of ineffective therapy.

• In MASLD and other liver diseases, genetic variants may improve biological understanding and eventually contribute to risk stratification.

• AI provides the computational framework needed to integrate large numbers of heterogeneous variables into patient-specific predictions.

• Personalized surveillance may eventually determine who needs screening, how often and using which modality.

• A useful precision model must demonstrate that acting on its prediction actually improves patient outcomes.

• Cost, accessibility and equity will determine whether precision medicine becomes routine care or remains concentrated in specialized centers.

Take-home: The destination is not simply more biomarkers—it is using the right information to make a different and better decision for an individual patient.

• Chronic GI disorders are episodic, while conventional outpatient assessment captures patients at only isolated time points.

• Digital platforms can continuously or periodically collect symptoms, medication adherence and patient-reported outcomes.

• Remote IBD monitoring can combine symptoms with objective biomarkers such as home or remotely coordinated fecal calprotectin testing.

• Wearables may contribute data on activity, sleep, heart rate and functional status.

• In cirrhosis, digital monitoring could potentially identify early deterioration in frailty, mobility or cognition.

• Telemedicine can improve access for patients living far from specialist centers and reduce unnecessary hospital visits.

• Automated alerts could trigger earlier clinical review when predetermined deterioration thresholds are reached.

• Digital systems should complement—not replace—clinical judgment and appropriate physical assessment.

• Important barriers include digital literacy, interoperability, data privacy, alert fatigue and uncertain reimbursement models.

Take-home: Digital medicine could transform follow-up from scheduled episodic visits to continuous, risk-triggered care.

• Modern endoscopy increasingly combines high-definition white-light imaging with image-enhancement technologies.

• Virtual chromoendoscopy can improve visualization of subtle vascular and mucosal patterns.

• Magnification and advanced imaging can support optical differentiation between neoplastic and non-neoplastic lesions.

• In Barrett’s esophagus and early gastric neoplasia, enhanced imaging can help direct targeted biopsy and therapy.

• Capsule endoscopy generates enormous numbers of images, making it particularly attractive for AI-assisted interpretation.

• Deep-learning systems can rapidly screen capsule studies for bleeding, ulcers and other abnormalities.

• AI may ultimately combine lesion detection, characterization and invasion-depth prediction within the same platform.

• Advanced imaging could reduce unnecessary biopsies when optical diagnosis reaches sufficiently high and reproducible accuracy.

• The major challenge is translating excellent performance on curated datasets into reliable real-time performance in routine clinical practice.

Take-home: The future is not simply better cameras—it is better imaging plus intelligent interpretation at the point of care.

• Microbiome research is shifting from cataloguing “dysbiosis” toward testing causal and therapeutic interventions.

• Recurrent C. difficile infection provides the strongest proof that microbiome restoration can produce major clinical benefit.

• Standardized microbiota-based therapeutics offer a more controlled approach than traditional donor-dependent FMT.

• In IBD, microbiome signatures may eventually contribute to phenotyping, prognosis and treatment-response prediction.

• The gut microbiome also interacts with metabolism, bile acids, immune signaling and the gut–liver axis.

• Interest is expanding into MASLD, IBS and other disorders, although evidence remains much less mature than for recurrent CDI.

• Future therapies may include defined bacterial consortia, engineered organisms, bacteriophages and targeted microbial metabolites.

• A major challenge is distinguishing causation from association—many microbial changes may simply reflect disease, medication or diet.

• Broad commercial microbiome testing currently exceeds what can be reliably translated into clinical decisions.

Take-home: Microbiome therapy is becoming real medicine, but outside established indications such as recurrent CDI, precision microbiome treatment remains largely investigational.

• EUS has evolved from a predominantly diagnostic modality into an increasingly therapeutic platform.

• EUS-guided tissue acquisition remains central to diagnosing pancreatic and other deep GI lesions, with improvements in needle design and sampling strategies.

• Contrast-enhanced EUS and elastography may provide additional information for lesion characterization and targeted sampling.

• EUS-guided drainage has transformed management of pancreatic fluid collections.

• EUS-guided biliary drainage is an increasingly important alternative when conventional ERCP is unsuccessful or impossible.

• EUS-guided gallbladder drainage offers a minimally invasive option for selected high-risk surgical patients.

• EUS-guided gastroenterostomy is emerging as an alternative to surgical or enteral-stent approaches in selected gastric-outlet obstruction.

• AI-based image interpretation could eventually assist with pancreatic lesion recognition and characterization.

• These procedures remain technically demanding and should be concentrated in centers with appropriate expertise, multidisciplinary support and rescue capability.

Take-home: EUS is transitioning from “look and biopsy” to “diagnose and treat”, increasingly bridging endoscopy, radiology and surgery.

• IBD now has numerous effective therapies, but choosing the right first therapy for an individual patient remains difficult.

• Clinical phenotype alone is insufficient to reliably predict response to biologics and small molecules.

• Precision medicine aims to integrate clinical characteristics, biomarkers, genetics, transcriptomics, proteomics and microbiome data.

• Fecal calprotectin and CRP already demonstrate the value of objective biomarker-driven monitoring over symptoms alone.

• Molecular signatures may eventually help predict which patients will respond to specific mechanisms of action.

• AI can potentially integrate complex datasets that are difficult for conventional clinical decision-making.

• Digital symptom monitoring may identify deterioration before a conventional clinic visit.

• Treatment should increasingly target deep disease control, not merely symptomatic improvement.

• At present, most multi-omic predictive models remain research tools rather than validated bedside treatment selectors.

Take-home: Precision IBD is moving toward the right drug, patient and timing—but robust prospective validation is still needed before multi-omics routinely chooses therapy.

• Hepatology is particularly suited to predictive analytics because longitudinal laboratory, imaging and clinical data are routinely generated.

• AI models can potentially improve identification of patients at risk of fibrosis progression and decompensation.

• In MASLD, combining routine variables with elastography, imaging and metabolic parameters may refine fibrosis-risk assessment.

• Longitudinal models may be more informative than a single FIB-4 or liver-stiffness measurement because trajectory matters.

• Predictive tools are being explored for HCC risk, potentially allowing surveillance intensity to become more individualized.

• In cirrhosis, algorithms may help predict decompensation, hospitalization, infection and mortality.

• Transplant medicine could benefit from models integrating donor, recipient and perioperative variables.

• Wearables and patient-generated data could add information on activity, frailty and functional deterioration.

• External validation remains essential: sophisticated algorithms can still fail when applied to different populations and healthcare systems.

Take-home: The future of hepatology may shift from asking “How advanced is the liver disease today?” to “What is likely to happen to this patient next?”

• AI can objectively measure aspects of colonoscopy that previously depended largely on manual reporting and auditing.

• Computer vision can assess withdrawal time, bowel preparation and mucosal visualization.

• Systems can identify blind spots and inadequately inspected colonic segments in real time.

• AI may quantify the percentage of mucosa adequately examined rather than relying only on total withdrawal time.

• Automated recognition of anatomical landmarks may improve documentation of cecal intubation and examination completeness.

• Quality systems can potentially combine ADR, withdrawal technique, preparation quality and inspection coverage into a procedural quality profile.

• This could provide immediate feedback for endoscopy training and competency assessment.

• AI-based auditing may help reduce operator-dependent variation between endoscopists and centers.

• Nevertheless, quality metrics should improve meaningful patient outcomes rather than create another layer of automated box-ticking.

Take-home: AI may evolve from finding polyps to monitoring how well the entire colonoscopy is performed.

• AI can contribute across the CRC pathway—from screening and lesion detection to characterization, pathology and risk prediction.

• Deep-learning CADe systems can highlight lesions during live colonoscopy, including subtle flat and small adenomas.

• CADx algorithms increasingly attempt to predict histology from endoscopic appearance.

• AI combined with high-definition imaging, NBI and virtual chromoendoscopy may improve optical characterization.

• Computer vision may help define lesion boundaries and identify features suggesting deep submucosal invasion.

• Digital pathology algorithms can assist with tumor detection, grading, segmentation and biomarker assessment.

• AI could eventually integrate endoscopic, histological, molecular and clinical data to improve personalized CRC risk stratification.

• However, algorithms trained in selected datasets may perform differently across populations, endoscopy platforms and real-world settings.

• The clinically important endpoint is not simply improved ADR—it is whether AI can reduce missed advanced lesions, post-colonoscopy CRC and unnecessary interventions.

Take-home: AI is moving CRC screening from simple lesion detection toward an integrated detection–diagnosis–risk prediction ecosystem.

• CADe is the most mature GI application of AI, providing real-time prompts for suspected colorectal lesions during colonoscopy.

• The principal demonstrated benefit is increased adenoma detection, particularly for small or subtle lesions.

• AI may reduce some of the variability in detection between high- and lower-performing endoscopists.

• The next step is CADx—characterizing detected lesions, potentially helping distinguish neoplastic from non-neoplastic polyps.

• AI-assisted optical diagnosis could eventually support “resect-and-discard” or “diagnose-and-leave” strategies in appropriately selected diminutive lesions.

• Beyond the colon, AI is being developed for Barrett’s neoplasia, gastric cancer and dysplasia, where subtle mucosal abnormalities are easily overlooked.

• A key limitation is false-positive alerts and alert fatigue, particularly from folds, bubbles and residual material.

• Higher detection does not automatically mean better outcomes; the critical question remains whether AI ultimately reduces interval cancer and cancer mortality.

Take-home: AI is increasingly becoming a second pair of eyes during endoscopy, but the goal should be clinically important detection—not simply more alerts.