UDR Fused Multimodal Approach for Disease Classification in Large Scale Dataset with Advanced CNNs

Breast cancer stands out as a major threat to women’s health, claiming a prominent position as the leading cause of female mortality. In this research, we present a holistic strategy for breast cancer multiclass disease classification employing advanced deep learning architectures, including a U-Net, DenseNet201, ResNet50 models referred as UDR fused multimodal. To optimize model training on large-scale datasets, we enhanced these models by integrating SVM components. As a result, we built an ensembled model consisting of U-Net, DenseNet201, and ResNet50 intricately combined these networks based on their weights and architectures, culminating in a highly accurate classification system with an accuracy of 91.43%. We conducted the extensive evaluations on the Histopathological Image Classification dataset utilizing deep learning methodologies, showcasing promised outcomes and underscoring the effectiveness of our fused UDR multimodal classification model. Our approach fused cutting-edge techniques, emphasizing transfer learning, and ensemble strategies to achieve superior performance on the Histopathological image classification dataset. Specifically, we explored various ensembled methods, including combining densely connected networks and leveraging pre-trained models for efficient training.