Improved convolutional architectures for breast cancer histopathological image analysis
Xile Zhu, Xiaomin Wang · 2025
Breast cancer remains the most prevalent malignancy among women globally, with substantial efforts dedicated to its early detection, diagnosis, and therapeutic intervention. This study proposes an enhanced deep learning framework based on classical convolutional networks (ResNet) and ConvNeXt architectures, evaluated on the publicly available Breast Cancer Histopathological Image Database (BreakHis). The experimental results demonstrated that our novel hybrid module not only strengthens feature extraction capabilities but also significantly improves recognition accuracy for complex pathological images. Compared with baseline ResNet architectures and other common ResNet variants, our proposed method exhibited superior performance across key metrics, achieving a 2.1% improvement in overall accuracy over the ConvNeXt baseline model. Particularly noteworthy enhancements were observed in recall rates, suggesting improved sensitivity in malignant tissue identification.