Oesophagogastric junction (OGJ) cancer represents a biologically heterogeneous malignancy located at the anatomical transition between the distal esophagus and proximal stomach, commonly categorized using the Siewert classification system. Despite advances in surgical techniques and multimodal therapy, prognosis remains poor due to late-stage diagnosis, tumor heterogeneity, and limitations in conventional imaging-based staging. Recent progress in artificial intelligence (AI), particularly deep learning and machine learning algorithms, has enabled the extraction of high-dimensional quantitative features from endoscopic and radiological images, giving rise to novel imaging-based biomarkers. AI-assisted endoscopy enhances early lesion detection, automated tumor margin delineation, and invasion depth prediction through real-time image analysis. Concurrently, radiomic analysis of CT, PET-CT, and MRI data provides quantitative assessments of tumor texture, shape, and heterogeneity, which can predict lymph node metastasis, response to neoadjuvant therapy, and overall survival. Integration of endoscopic biomarkers, radiomic features, and molecular profiles—termed radiogenomics—further advances precision oncology by enabling personalized risk stratification and treatment planning. Although promising, challenges such as data standardization, reproducibility, interpretability, and regulatory validation remain significant. This review synthesizes current evidence on AI-integrated endoscopic and radiomic biomarkers in OGJ cancer and discusses their translational potential toward improving diagnostic accuracy, prognostic prediction, and individualized therapeutic decision-making.
Acharya et al. (Thu,) studied this question.