DEVELOPING A BILINGUAL SELF-LEARNING MEDICAL KNOWLEDGE BASED SYSTEM FOR DIAGONOSES, AND TREATMENT OF MALNUTRITION CAUSED DISEASES FOR CHILDREN

Abstract

Malnutrition is unbalanced intake of foods which causes kwashiorkor, marasmus and diarrhea in children under 5 years old. Kwashiorkor and marasmus have almost similar infection in children, while kwashiorkor is deficiency of protein, marasmus is deficiency of protein and calories, and diarrhea on the other hand is caused by pollinated and low intake. Therefore, having standard children feeding system is crucial to prevent such diseases. In Tigray; children under the age of five year are stunned and caused death because of malnutrition diseases. Child mortality rate was 12 deaths per 1,000 children surviving to the age of 12 months, while the overall under-5 mortality rate was 55 deaths per 1,000 live births. The societies have no more awareness of the malnutrition system. In addition, there is a lack of enough health centers and experts especially in 2021 in the domain area. Considering such problem, this study attempts to design and develop a bilingual prototype self-learning knowledge based system that can provide advices and treatments on malnutrition caused disease such as kwashiorkor, marasmus and diarrhea. It also suggests standard feeding system for children. To do this, researcher acquired knowledge using structured and unstructured interview from domain experts and relevant documents from Ayder Referral Hospital Mekelle. After that, acquired knowledge was modeled using decision tree that describe the producers of diagnose and treatment of the disease. The modeled knowledge was represented using production rule and SWI Prolog editor tool used to develop the system. To verify the facts and propose solution, backward chaining method was followed for reasoning, which is a goal driven approach. The system performance and user acceptance testing is performed by selecting twenty patients cases purposively. This ensures to measure the satisfaction level of users and usability of the prototype. Prototype registered 90.4% and 89.9% system performance and user acceptance respectively. On the average, the performance of the prototype is 90.15%. The result shows that study achieved a promising result. The system should learn from domain experts so as to know new facts and their relationship followed by self-learning. There is a need therefore to integrate incremental learning to enable the system learn through feedback received from experts