Evidence record · UDINUS · SDGs
- Unit
- Rektorat UDINUS
- Period
- 2026
- Goals
- Record
- a686f3ec-7dba-4dd1-86c4-910c328d6e11
- Published
- 2026-10-30
Universitas Dian Nuswantoro conducted an activity on breast cancer classification using histopathological images through the development of the Breast Advanced Swin Transformer Network, or BreAST-Net. The activity involved Nugroho, Wahyu Adi; Supriyanto, Catur; Pujiono; Wijaya, Adi; and Alomoush, Ashraf as participants. The activity addressed breast cancer, which is described as among the foremost causes of mortality among women in Indonesia, with both incidence and mortality rates rising annually. It also addressed the importance of timely detection for identifying suitable treatments according to the specific type and subtype of cancer and for decreasing associated mortality rates.
The activity examined deep learning methodologies for medical image classification. The source explains that Convolutional Neural Networks have limitations associated with localized feature extraction. This limitation hinders their ability to capture global interactions among cells in histopathological images. BreAST-Net was developed as a model based on the Swin Transformer, with adjustments to the global activation function. The standard Gaussian Error Linear Unit activation function was substituted as part of this adjustment.
Experiments were conducted using the BreakHis dataset. BreAST-Net was compared with CNN-based models and the conventional Swin Transformer. The model demonstrated improved performance relative to those approaches. For breast cancer type classification, the reported highest accuracy, precision, recall, and F1-score were 99.60%, 99.50%, 99.60%, and 99.55%, respectively. For breast cancer subtype classification, the corresponding values were 97.57%, 97.04%, 97.57%, and 97.24%. The results suggest that BreAST-Net may be integrated into clinical diagnostic systems as an accurate and efficient method for early breast cancer diagnosis. The activity therefore presents UDINUS research addressing breast cancer detection and classification through a deep learning approach.