SOURCE:

Faculty: Physical Sciences
Department: Computer Science

CONTRIBUTORS:

Enechi, G. O.
Inyiama, H. C.

ABSTRACT:

A number of avoidable blindness could have been averted if efficient eye diagnostic system were available and accessible to persons suffering from various eye diseases especially in remote and inaccessible communities where there is no eye doctor or where the eye doctors are reluctant to go. The existing eye diagnostic model uses Bayesian Neural Network which is less efficient in making decisions involving small databases that are mostly prevalent in eye care medicine. It neither uses visual acuity of patient as validating tool for patient’s symptoms nor allows the patient to provide the level of confidence associated with his/her symptoms. In addition, the existing system does not integrate diagnosis and visual acuity measurement into a single platform. Development of a model that can surmount the setbacks prevalent in the existing system and at the same time provide reliable eye disease diagnosis, is sine qua non for prevention of avoidable blindness especially in developing countries. This dissertation is aimed at developing a Certainty Factor Reasoning Model for eye disease diagnosis. Rule Based Expert System Methodology and Object-Oriented Analysis and Design Methodology were used for knowledge and data representation respectively. Certainty Factor Method was used for reasoning under uncertainty. Visual Prolog 7.5 was used for the design and implementation of the new system while MYSQL 1.0.9 was used for the design and implementation of the database. Disease diagnosis was inferred by analyzing the visual acuity, the symptoms and certainty factor provided by the patient with the facts, rules and also the certainty factor provided by the domain expert. Test of accuracy for the diagnosis made by the new system using Cohen’s kappa tools revealed k-value of 0.827 while Test of accuracy and test/retest reliability for the visual acuity measurement using Bland-Altman tools revealed p-values of 0.042 and 0.00004 respectively. The stress/overload test was successfully carried out on 154 operations. These performance indicators imply that the new system is accurate, reliable and repeatable. This study has reinforced the belief that certainty factor model could be a useful reasoning tool for making decisions especially in uncertain situations and in small databases. It is believed that the model could help in eradication of avoidable blindness as well as provide training tools to eye doctors, clinical students and can also serve as information portal to the patients. Further research into development of web-enabled version of this model to extend its reach is recommended.