Classification of Biological Phenomenon-of-Interest via Dielectric Information Probe

This paper proposes a novel framework to classify $in vivo$ biological phenomenon-of-interest (BPI) based on a new multi-stage biological information transmission model. Specifically, the classification is realized by exploring the statistical relations between three interconnected blocks: the BPI that corresponds to the pathological condition under classification, the dielectric information probe (DIP) that exhibits BPI-sensitive dielectric properties, and the external observation system (EOS) that interprets the features of the DIP. Consequently, optimised probabilistic mappings can be identified among these blocks to ensure effective classifications. This framework is illustrated by identifying a non-invasive breast tumor based on the percentage of cancerous tissues that forms glands with contrast-enhanced microwave imaging. By analyzing the statistical relations among the above three blocks for this specific problem, the accuracy of identifying an invasive tumor could reach 90% in a simulation setting.