Hybrid Case Based Reasoning Approach for Diagnosis and Treatment of Hypertension

Abstract

Health problems touch many aspects of human life like health condition, working environment, family life, social relations, financial and political activities of every endeavor. In Ethiopia, every sector has been adversely affected by a variety of chronic diseases like hypertension. Hence, the effective use of Knowledge base system in providing health services is one of the most essential approaches for addressing shortage of qualified health professionals, experts, advisers and trainers in the area. Therefore, this study strongly attempts to investigate hybrid CBR approach by integration of ANN (Artificial Neural Networks) with CBR (Case Based Reasoning) to provide Diagnosis and Treatment of hypertension. The proposed knowledge based system is implemented to leverage the advantages that this combination brings an improvement in effectiveness of CBR retrieval and similarity measure. The study followed the design science research approach with six steps process model. For problem identification and formulate the objective of the solution, knowledge is acquired by using document analysis, domain expert interviewing and previously solved patient cards. The required data were collected from South West region of Ethiopia, particularly from Bonga Gebretsadik Shawo General Hospital, Wacha First Level Hospital, and Bonga Health Center. Likewise, the domain knowledge was acquired from domain experts who work in the same health facilities. Once the data collection task was completed, the dataset was prepared experimentation. After implementation of the CBR and ANN independently, the proposed hybrid CBR approach is implemented subsequently. Following it, experimentations were conducted to test and evaluate the proposed prototype based on test cases and validated by the domain experts. The result shows that the prototype passed all the test cases. Finally, the proposed hybrid CBR system is developed with graphical user interface as a web application in English and Kafa languages using Flask framework. Two web application is developed, one for English language users and the other is for the local language Kafa users. After the prototype was demonistrated for domain experts, they rated on average 4.13 out of 5, which is equivalent to the Likert scale “Very Good” that is highly encouraging. Finally, the future work needs to examine the significance of weighted input attributes to improve the performance of the proposed system as it would mimic the evaluation behavior of human based methods.