Bayesian network construction using a fuzzy rule based approach for medical decision support

Abstract

This study proposes a novel method for the construction of efficient and convenient Bayesian networks (BNs) and influence diagrams regarding medical problems based on fuzzy rules. The general methodology that was developed is able to address decisions based on fuzzy medical rules that connect symptoms with the severity/rating scale of a disease. These fuzzy rules are rich enough to cover a large variety of medical decisions. The method overcomes the major disadvantage of Bayesian nets, that is, the need of a vast amount of subjective probabilities. Instead of filling conditional probability tables in BNs, physicians report their knowledge in the form of fuzzy rules. The knowledge of these rules after defuzzification is transformed according to certain equations into probabilities. This becomes possible, categorizing the rules into certain types. For each type of rule, a mathematical expression is determined, which sets correctly the conditional probabilities that relate a symptom with its child severity. A particular example of assessing pulmonary infections and making decisions on severity degree was explored, implementing a decision support system (DSS) to show the functionality of the proposed methodology. The developed front-end DSS was evaluated and proved capable to drive fair decisions close to those obtained by physicians.

Knowledge discovery using maximization of the spread of influence in an expert system

“…The design of ESs using a conceptual approach has been reported in the past (Amirkhani et al, ; Clancey, ; Swaffield & Knight, ; Zarikas, Papageorgiou, & Regner, ). Some of the most important issues that have arisen in this article were already discussed in the literature, including the development of structured reusable libraries.…”

A multiperiod hybrid decision support model for medical diagnosis and treatment based on similarities and three‐way decision theory

“…A FS A in the domain space X can be characterized by the membership function μ A (x) : x ∈ X → [0, 1]. FS has been applied to various research areas, such as replenishment system (Leung, Lau, & Kwong, 2003) and medical diagnosis (Binaghi, 1990;Zarikas et al, 2015). Atanassov (1986) proposed IFS to generalize the ordinary FS concept.…”

“…Medical diagnosis is considered to be a convenient support tool in clinical medicine that helps physicians to determine the most possible disease and give appropriate medicated figures on the basis of a set of given symptoms. In the last few years, numerous approaches have been introduced to address medical diagnosis problems in an efficient way, including learning machine (Gürbüz & Kılıç, 2014;Qiang, Ge, Zhao, Zhang, & Tang, 2017), case-based reasoning (Chattopadhyay, Banerjee, Rabhi, & Acharya, 2013;Park, Kim, & Chun, 2006), Bayesian network (Zarikas, Papageorgiou, & Regner, 2015), and statistical or pattern recognition methods (Hemanth, Anitha, & Ane, 2017;Wolfers, Buitelaar, Beckmann, Franke, & Marquand, 2015), among others. Nonetheless, a crucial issue in medical practice is that patients' original information is usually imprecise and uncertain because the collection of information is expensive, time-consuming, and even harmful to patients.…”

A methodology for assessing the effect of portfolio management on NPD performance based on Bayesian network scenarios

“…Hou, Zhao, Zhao, and Zhang (2016) used dynamic Bayesian networks to predict mobile users' behaviours and locations. A medical decision support system is also developed based on BNs to assess pulmonary infections and to make decisions on severity degree (Zarikas, Papageorgiou, & Regner, 2015).…”