Perspective: Microwave Medical Imaging Using Space-Time-Frequency A Priori Knowledge for Health Monitoring

Abstract

Microwave medical imaging (MMI) operating over the frequency range covering hundreds of megahertz to tens of gigahertz has the potential to provide proactive healthcare solutions to patients with acute (for early diagnosis) or chronic (for daily monitoring) medical conditions. This technology exploits the tissue dielectric properties for disease diagnosis by using quantitative or qualitative algorithms. The advantages of MMI include low health risk, low operational cost, lightweight implementation, and ease of use, given its perspective of miniaturization and integration into portable and handheld devices with networking capability. MMI has been proposed for cancer detection, stroke detection, heart imaging, bone imaging, tracking of in-body drug-loaded nanorobots, etc. It is, however, challenging to develop accurate and robust MMI algorithms for both sensitive and selective diagnosis, due to the inherently ill-conditioned inverse scattering problems and the low dielectric contrast between healthy and diseased tissues. As such, using the a priori knowledge (APK) about the scattering profile to improve the performance of MMI is crucial for practical implementation and clinical deployment of MMI systems. This perspective article presents a new viewpoint of categorizing and utilizing various types of APK, which is acquired from the space, time, or frequency (STF) domain. The article starts with a general categorization framework of APK, followed by formulations of MMI algorithms utilizing APK. Subsequently, the existing APK-oriented MMI algorithms are reviewed in the respective STF domain. Finally, the influence of accuracy of APK on MMI performance is discussed using numerical examples. Through the analysis of the distorted Born iterative method (DBIM) and the pulse radar method, we have discussed the accurate usage of time-domain APK for both quantitative and qualitative evaluations, and the performance improvements of the quantitative and qualitative algorithms are 92% and 80%, respectively. The results demonstrate that the proper implementation of APK can significantly improve imaging accuracy, further validating the effectiveness and generalizability of the proposed model. This perspective would offer some useful insights into the future directions of MMI algorithmic development.