Academic journal article Communications of the IIMA

Composite Ontology-Based Medical Diagnosis Decision Support System Framework

Academic journal article Communications of the IIMA

Composite Ontology-Based Medical Diagnosis Decision Support System Framework

Article excerpt

INTRODUCTION

Medical decision making covers important tasks such as diagnosis, therapy planning, interacting with patients, identifying medical errors etc. Medical diagnosis is a process aiming at identifying diseases based on findings, such as symptoms and lab reports. The development of Medical (Clinical) Diagnosis Decision Support Systems (MDDS or CDSS) dates back to 1950s. Such developments, particularly in diagnosis decision support, have high complexity. A limited number of systems are adopted for practical use in the clinical environment. In the diagnosis process, an appropriate representation scheme is necessary for both problem interpretation and knowledge retrieval. From a medical cognition point of view, Long (2001) states that most medical reasoning methods are based on organizing various types of relations that exist in medical domain. These relations were identified by Long (2001) and include: associations, probabilities, causality, functional relationships, temporal relations, locality, similarity, and clinical practice (cases and experiences).

In computer science community, early diagnosis decision support (appeared as expert systems (ES)) research focused on rule-based reasoning (RBR) methods, decision table/tree, and later on Bayesian probabilistic, case-based reasoning (CBR). More recently as the computational power improves, machine learning-based systems emerged. These efforts were novel at the time; however, they did not employ a formal knowledge model (e.g., model of diseases, causation etc.), instead, they relied on the characteristics of data. Further, these approaches mostly operate in syntactic manner. Therefore, it is typically difficult to generate semantic explanation to the decision made by the system. The evolving Semantic Web research has brought a new platform for better knowledge representation, its sharing and semantic reasoning. The ontological model now serves as building blocks for the representation tasks in most knowledge-based applications.

Essentially, a good medical diagnosis system requires a structured knowledge representation component (model) that reflects most of the existing medical relations. It also needs employing efficient reasoning methods that closely follow medical cognition. Organizing these relations obviously needs one or more specific forms of knowledge representation, and computational artifacts that can manipulate them. None of these tasks is easy. Furthermore, physician-like users usually lack knowledge of how the framework works. In the worst case, a common bottleneck in knowledge-based systems is the knowledge acquisition, because of the acquisition mechanism is not transparent to the experts, or tacit knowledge is hardly complete enough to be usable.

We explore the case-based reasoning (CBR) methodology, as CBR not only utilizes the actual data (cases), but also works similarly to how human solves problems by recalling most relevant experiences. We propose a framework for medical diagnosis decision making. This framework incorporates the disease-symptom ontology, and case-based reasoning (CBR) coupling with semantic similarity calculation.

The paper is organized as follows: the next section introduces the readers to knowledge representation. The third section summarizes BioMedical ontologies and the semantic web technologies. The fourth section explains how CBR works. The fifth section presents our composite ontology framework. The diagnosis workflow is set up in the sixth section. This is followed by the conclusion.

KNOWLEDGE REPRESENTATION

In problem solving, knowledge provides the basis for reasoning in either a modal or an ad hoc way. Researches in artificial intelligence early on aimed at using computer power to act as human intelligence for problem solving. A human problem solver either exploits his own or others' experiences (if he understands well). Alternatively, he may visit available formal models (e. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.