Breast Cancer Classification Based on Histopathological Images Using a Deep Learning Capsule Network
This work addresses the problem of accurate breast cancer diagnosis for clinicians and patients, but it appears incremental as it builds upon existing capsule network architectures.
This study tackled breast cancer classification from histopathological images by proposing an enhanced capsule network with Res2Net blocks and additional convolutional layers, achieving improved performance over previous deep learning methods.
Breast cancer is one of the most serious types of cancer that can occur in women. The automatic diagnosis of breast cancer by analyzing histological images (HIs) is important for patients and their prognosis. The classification of HIs provides clinicians with an accurate understanding of diseases and allows them to treat patients more efficiently. Deep learning (DL) approaches have been successfully employed in a variety of fields, particularly medical imaging, due to their capacity to extract features automatically. This study aims to classify different types of breast cancer using HIs. In this research, we present an enhanced capsule network that extracts multi-scale features using the Res2Net block and four additional convolutional layers. Furthermore, the proposed method has fewer parameters due to using small convolutional kernels and the Res2Net block. As a result, the new method outperforms the old ones since it automatically learns the best possible features. The testing results show that the model outperformed the previous DL methods.