DMRBNet: Dilated Multi-Scale Residual Block-Based Deep Network for Detection of Breast Cancer from MRI Images

Breast cancer (BC) is a common type of cancer that develops from breast tissue cells. Early detection is critical, and mammography is an important tool for this. A biopsy is indicated for lesions with a risk of malignancy of more than 2%, however, only a tiny number of them are confirmed to be malignant. Magnetic Resonance Imaging (MRI) is employed to eliminate unneeded biopsies, but it is a sophisticated and time-consuming operation requiring specialized knowledge. To improve breast cancer diagnosis, a computer-aided diagnostic system using MRI images was developed. The system utilizes a novel neural network called dilated multi-scale residual block-based convolutional neural network (DMRBNet), which effectively extracts features from various image regions. Compared to seven recent advanced approaches, DMRBNet demonstrated superior performance on the BC-MRI dataset. The accuracy of the network is 98.57%, and the error rate is 0.1005. These findings highlight its potential for medical and industrial applications in breast cancer detection.