BCViMer 1.0: A Deep Network Pipeline for the Identification of Breast Cancer Using MRI Data

Breast cancer ranks among the foremost causes of cancer-related deaths in women globally, with early diagnosis via advanced imaging techniques significantly enhancing patient outcomes. Convolutional neural networks (CNNs) demonstrate considerable potential in the automation of breast cancer detection. Nonetheless, their constrained capacity to capture global image features may hinder their performance. This study presents BCViMer 1.0, a deep learning pipeline utilizing the Vision Transformer (ViT) to overcome existing limitations and enhance breast cancer classification accuracy from MRI images. This research aimed to assess the efficacy of BCViMer 1.0 relative to established CNN architectures, specifically ResNet-50, ResNet-101, VGG-16, and VGG-19. Experimental results indicate that BCViMer 1.0 excels these CNN models, attaining an average classification accuracy that is 4.15% higher. The enhanced performance of BCViMer 1.0 is due to its capacity to capture long-range dependencies and global context in MRI images, facilitated by the self-attention mechanism of the ViT. In conclusion, BCViMer 1.0 exhibits superior diagnostic accuracy compared to conventional CNN-based models, representing a significant advancement in breast cancer detection. This study emphasizes the capabilities of Vision Transformers in medical imaging and establishes a basis for future investigation in clinical applications.