الفهرس 
AbstractOnly 14 pages are availabe for public view
 |  | from | 286 |  |  |
| | | from | 286 | | |
Abstract
Diabetes is one of the most serious chronic diseases. It is the seventh leading cause of death. It imposes a large economic burden on the individual, national healthcare systems, and countries. How to decrease its threats is a challenge. The early diagnosis is the first and most critical step in diabetes management process because it can prevent its long-term microvascular complications as retinopathy, nephropathy and neuropathy, and cardiovascular diseases. There are many clinical practice guidelines for standardizing the diagnosis process; however, these guidelines are long text documents, which are difficult to be used by physician at the point of care. Integrating Clinical Decision Support Systems (CDSS) in the Electronic Health Record (EHR) systems at the point of care can improve the solution of this problem. CDSS can provide the right physician, with the right information, in the right form, at the right time. Case Based Reasoning (CBR) is one of the most suitable techniques for solving these types of problems. This research adds an advanced step in the implementation of CDSSs. The study proposes and implements a complete framework for fuzzy-semantic CBR system. The framework has many phases that are completely implemented and tested. The diabetes diagnosis is the studied and tested medical problem. We have solved most of the current issues that prevent the execution and utilization of CBR-based CDSS in the health care environments. CBR has a standard lifecycle with five phases namely case-base elaboration, case retrieval, case reuse, case revise, and case retention. The first two phases are the most critical steps that determine the whole system performance, and we concentrate on these two steps. The first is the building of CBR system’s knowledge base (i.e., case-base or experience-base), and the second is case retrieval. Regarding the first phase, we have proposed a standard case-base data model based on HL7’s Reference Information Model (RIM). This data model collects the diabetes elements from the most recent clinical practice guidelines, our domain experts, and recent diabetes researches. Moreover, we have proposed a case-base preparation framework to extract patients’ records from EHR and convert it to knowledge cases; this framework has three main steps namely data pre-processing, data fuzzification, and data coding. Regarding the second phase, we have proposed a hybrid case retrieval algorithm based on the case-base contents including numerical, categorical, fuzzy, ordinal, semantic, and fuzzy-semantic. The proposed algorithm is comprehensive and implemented in java. For testing our work, we have collected a list of 60 real diabetic patients from Mansura University Hospitals EHRs. Each step of our framework has been tested separately and in conjunction with other steps. The cases are prepared using our framework. The pre-processed, fuzzified, and encoded patient’s data are populated in the case-base ontology. We have tested our knowledge base and retrieval algorithm using the leave-one-out methodology. The proposed framework achieved a high performance level of precision= 100%, recall=96.43%, accuracy=97.67%, specificity=100%, F-measure= 98.18%. These results show that the proposed system has a high accuracy, and physicians can consult it when diagnosing patients.