Reasoning methodologies in clinical decision support systems: A literature review

Abstract

Abstract The growing significance of clinical decision support systems (CDSS) has become a positive factor influential in pushing medical care toward success. It depends on using successful and effective reasoning methodologies. This survey aims to give a brief overview of the research directions that are practiced under the domain of reasoning methodologies used in CDSS implementation. It focuses on studying the roles of fuzzy ontology and fuzzy logic in the CDSS implementation given in the scientific literature. We are trying to identify new future trends in this domain. We adopt a search methodology involving the definition of research questions, the determination of selection criteria, and the description of the search strategy. The primary questions of this review are as follows: Which reasoning techniques have been used in CDSS? What is the accuracy of using different reasoning techniques in real applications? What are the limitations of existing reasoning techniques? How to enhance the reasoning process in DSS? The manuscript describes the current published literature in Science Direct, Springer Link, PubMed, and IEEE Xplore from 2009 through November 2017. The search strategy contains four processes: screening papers, selecting papers, extracting and analyzing concepts, and identifying future trends. Our search identified 1886 papers across different electronic databases. These papers are used as an initial database. After reviewing these articles, we selected 134 relevant articles that are more interesting and suitable for the goals of this paper. These relevant articles are included in our critical analysis to find the possible future trends. The literature review showed that case-based reasoning (CBR), Mamdani fuzzy inference, and ontology systems are the most-used reasoning techniques. However, the fuzzy inference failures, the unclear and not unified methods for the fuzzy ontology construction process and tools, the limitations of existing fuzzy description logic reasoners, and the manual case adaptation process in CBR are still the main problems and might not support the clinical practice effectively. Most of these models used ontology and fuzzy logic as two separate models, and no real overlap occurs. There are some serious points to be discussed to enhance the inference of the fuzzy component. Ontology can be used to enhance the capabilities of the fuzzy inference system. Our solution is the hybridization of regular and mature crisp ontology reasoning with regular and mature Mamdani fuzzy reasoning. We expect that will be the best choice to overcome the current limitations of crisp ontology and fuzzy reasoning. In this paper, the different reasoning methodologies applied to CDSS are analyzed. We are looking to combine ontology and Mamdani fuzzy inference in a hybrid CDSS system. The hybrid model is the most logical step to improve the fuzzy expert system by adding a semantic reasoning process to its capabilities. There are many reasons for this decision. First, the fuzzy expert systems are stable and mathematically proved, and there are many fuzzy reasoners such as Mamdani, etc. In addition, crisp ontology and its special case of (standard) medical ontologies have stable, crisp description logic such as SROIQ D , well-known languages such as OWL 2, and well-established reasoners.