Canonical Fuzzy Modeling of Disease State.

We propose a new theoretical framework for quantification and sensitization of disease state (DS). This is in contrast to the traditional discrete description of disease, where the inherent characteristic of its progression is ignored, and a finite number of DSs are resulted, effectively leading the sensing process to the act of classifying. Central to the framework is a canonical fuzzy model of DS that allows for conversion of its linguistic description into a normalized numerical variable. This generates the mapping between one set with the domain of the universe of DSs and another set with the domain of the universe of disease labels (DLs). Subsequently, the framework is analyzed from the fuzzy-set-theoretic perspective by utilizing the medical expert knowledge system of disease severity, consisting of the Acute Physiology and Chronic Health Evaluation (APACHE), which provides useful insight that enables the disease diagnosis process to be designed and optimized as a soft computing problem. Built on this fuzzy analysis, we present two multimodal strategies, namely the reversibility combining (RVC) and reliability combining (RLC), to enhance the sensing performance measured through the degree of non-reversibility of the surjective mapping from DLs to sensing outputs (SOs). Finally, we utilize some numerical examples based on highly realistic synthetic medical data to elaborate on the proposed framework, which offers a new perspective for disease diagnosis by monitoring continuously the individual's disease progression.