Document resource
Ovarian cancer (OC) remains a lethal gynaecologic malignancy characterised by late diagnosis, heterogeneity and limited stratification. While artificial intelligence (AI) has shown promise in pathology, single-modal approaches often fail to capture the interplay between morphological, molecular and clinical data. Multimodal AI (MMAI) integrates histopathological images, genomics, proteomics, radiology and clinical variables to generate holistic tumour profiles. This review synthesises recent MMAI advances in OC through three themes: methodological foundations, fusion strategies and clinical applications. Methodologically, MMAI leverages deep learning-based feature extraction and fusion architectures to align heterogeneous data. Clinically, MMAI enhances diagnostic accuracy, predicts therapy response (eg, platinum agents, poly (ADP-ribose) polymerase inhibitors), improves prognostic stratification and infers molecular features from histology. Challenges in data heterogeneity, interpretability and clinical integration persist. Future directions include self-supervised learning, federated training and spatial multi-omics to improve generalisability and biological insight. By providing a more comprehensive view of tumour biology, MMAI optimises therapeutic strategies and advances managing OC as a chronic disease.