Efficient Multi-Scale Attention Residual Network for Breast Cancer Histopathology Image Classification
Lu Cao, Ke Pan · 2023
In recent years, convolutional neural networks have achieved significant success in the field of breast cancer histopathological image classification. This study reexamines the application of scale analysis and cross-spatial learning, introducing EMAResNet, a novel breast cancer classification approach based on multi-scale attention mechanisms. EMAResNet integrates feature grouping and cross-spatial learning, efficiently processing and distributing spatial semantic features while significantly reducing computational complexity. By parallel processing of multi-scale features and efficient feature fusion, the model effectively captures and utilizes crucial information in image data. This approach has been extensively evaluated on the BreakHis breast cancer pathology image dataset, demonstrating superior classification performance.