S656 AI-Augmented Surveillance in Gastrointestinal Oncology: Enhancing Recurrence Prediction and Risk-Adaptive Monitoring

Introduction: Gastrointestinal (GI) cancers carry a high risk of recurrence post-treatment, yet conventional surveillance protocols are often non-adaptive and delay detection. This study evaluates the integration of artificial intelligence (AI) to enhance risk stratification and tailor follow-up regimens. Methods: We performed a systematic comparative analysis of 23 studies (2015–2024) to compare AI models (neural networks, radiomics) with traditional clinician-based risk assessments for recurrence prediction. Outcomes included predictive accuracy (AUC), early detection (months), false-negative rates, and recurrence underdetection. Sub-analysis incorporated molecular biomarkers (MSI, KRAS) and staging data. Results: AI models significantly outperformed clinician assessment in accuracy (AUC: 0.88 vs 0.69) and identified recurrence 2.4 months earlier on average. False-negative rates were reduced by more than 50% (6.1% vs 15.4%). AI-driven surveillance schedules were more intensive for high-risk patients and contributed to a 25% reduction in delayed recurrence detection. Biomarker integration further improved model specificity. Conclusion: AI-guided post-treatment surveillance for GI cancers enables dynamic, personalized monitoring that outperforms conventional methods. Integration into clinical workflows may improve early detection, optimize resource use, and enhance survival outcomes. These findings support the development of AI-guided clinical pathways for recurrence surveillance in GI oncology.