Bridging Knowledge Gaps in Multimodal Breast Cancer Diagnosis Using CTGA-Net

Breast cancer remains the most frequently diagnosed malignancy among women worldwide, with early detection significantly improving survival outcomes. Although mammography, ultrasound, and histopathology provide complementary diagnostic insights, existing artificial intelligence solutions largely remain unimodal and lack standardized multimodal fusion strategies. This paper proposes a Multimodal Convolutional Neural Network (CNN)–Transformer–Bidirectional Gated Recurrent Unit (BiGRU) with Attention Network (Multimodal CTGA-Net) for comprehensive breast cancer diagnosis across mammogram, ultrasound, and histopathology images. The architecture integrates CNNs for local feature extraction, Transformers for modeling global dependencies, BiGRU for contextual learning, and attention mechanisms for discriminative feature refinement. A cross-modal attention fusion module is introduced to address feature distribution mismatch and domain shift among heterogeneous modalities. The framework aims to bridge knowledge gaps including unimodal limitations, absence of unified segmentation–classification pipelines, and limited interpretability. Experimental design, architectural formulation, and expected clinical contributions are presented. The proposed approach provides a scalable and explainable AI-driven decision-support system adaptable to low-resource healthcare environments.