EMiTD: Enhanced Microwave Imaging for Breast Tumor Detection

In this paper, we propose an Enhanced Microwave imaging technique for efficient breast Tumor Detection referred as EMiTD, which mainly comprises of two key steps: an intelligent scanning approach to optimize the scan duration and an efficient model-based microwave imaging (MWI) technique. The proposed model-based MWI is framed as an inverse problem by building the forward model using a Point Spread Function(PSF), and is solved by imposing sparsity prior since tumor is concentrated to limited regions. Further, we optimize the entire scanning duration by viewing the problem as a sequential decision making process for a Deep Reinforcement Learning (DRL) agent. We benchmark the proposed EMiTD against other competing techniques using a publicly available dataset. Both visual and quantitative reconstruction results are provided which indicate that the proposed EMiTD as compared to other competing techniques provides significantly improved tumor localization with close to 2 x improvement in Signal to Mean Ratio (SMR). Furthermore, negligible deterioration in visual results is observed with the proposed approach in spite of using only 33 % of the total measurements.