A Systematic Pipeline for Evaluating Factors Affecting Radar-Based Microwave Breast Imaging Quality

Microwave imaging (MWI) has emerged as a promising non-ionizing, cost-effective technique for breast cancer detection. However, the influence of key system design choices remains poorly characterized. This study aims to systematically investigate how antenna array geometry, tumor characteristics, and tissue dielectric properties influence radar-based image reconstruction using a robust computational pipeline. A total of 575 different scenarios were simulated and analyzed, combining 25 distinct breast models with varying tumor configurations and 23 antenna configurations (including 20 randomly selected antenna subsets and 3 uniform arrays). The breast models included one or two tumors with varying radii, embedded in backgrounds with different dielectric properties. Quantitative metrics such as signal-to-noise ratio (SNR), signal-to-mean ratio (SMR), localization error, and separation error were used to consistently evaluate the reconstructed images. Additionally, a proof-of-concept experimental study using a laboratory MWI prototype was also conducted to demonstrate the applicability of the proposed pipeline to measured data. Simulation results show that image quality improves with increasing antenna count, while, among arrays with the same number of elements, uniform and symmetric geometries achieve higher average SNR than asymmetric configurations. Tissue relative permittivity and tumor size show limited correlation with the image quality metrics, whereas tumor location within the phantom significantly affects localization accuracy. The experimental proof of concept further demonstrated the capability of the proposed pipeline to compare different types of reconstruction algorithms using the same quantitative image quality metrics. These findings provide quantitative insights into the influence of antenna array design on breast MWI performance and demonstrate the value of systematic simulation-based evaluation frameworks. The proposed methodology can support the optimization of antenna array layouts and facilitate the development of more effective MWI systems prior to clinical translation.