A Deep Learning Framework for Breast Cancer Detection from RF Microwave Data

Breast Microwave Imaging (BMI) has emerged as a promising alternative to traditional breast cancer screening methods due to its non-ionizing nature and cost-effectiveness. This work proposes a deep learning-based BMI framework. A novel customized convolutional neural network (CNN) is proposed. In addition, for clinical practicality, this paper proposes replacing full-band scanning data with partial data, which can significantly reduce the scanning time while maintaining high accuracy. Experimental results show that the proposed deep learning model achieves superior performance in terms of detection precision, recall, accuracy, and F1-score. Meanwhile, it reduces scanning time by a factor of 3.3-10, with only a minor performance trade-off, demonstrating its potential to enhance BMI's clinical practicality.