Multi-Modal Feature Fusion Framework for Ovarian Cancer Classification

Accurate classification of Ovarian cancer subtypes is essential for improving diagnostic precision and personalised treatment strategies, as currently no effective screening test is available for this disease. Multi-Omics data offer complementary biological insights. However, effectively integrating these high-dimensional and heterogeneous datasets remains a major challenge. In this study, we propose CoWeMI, a novel multi-omics hybrid classification model designed to enhance Ovarian cancer subtype prediction through structured integration and weighted fusion of Copy Number Variation, DNA Methylation, miRNA expression, and mRNA expression data. The architecture of the proposed CoWeMI 3-layer model consists of : (i) Omics Data Integration layer, (ii) Hybrid Fusion layer to form a single fused feature matrix of 284 samples × 120 weighted features, and (iii) Classification layer. Experimental evaluation demonstrates that CoWeMI consistently outperforms the five machine learning models, Random Forest, Support Vector Machine, Logistic Regression, Gradient Boosting, and K-Nearest Neighbours under stratified 10-fold cross-validation, achieving a higher classification accuracy of 81.70% accuracy, 84.23% macro precision, and 0.9508 macro AUC with Random Forest and improved stability across multiple validation settings. These results highlight the potential of structured multi-omics integration and hybrid learning strategies to advance the precision of detecting Ovarian cancer. Keywords—Ovarian Cancer, Multi-omics, Classification, data integration, artificial intelligence, Gene expression