Diabetes mellitus diagnostic expert system

Abstract

The most challenging problem troubling the world today is spread of Diseases and the consequent sufferings of the innocent inhabitants of this planet. The doctors and researchers all over the world work, jointly and separately, day and night to combat this grave problem to the maximum extent they can. Inspire of this conscious and collective effort mistakes are due to wrong diagnostics, the starting step toward implementing a cure. Engineer can contribute in this filed by developing a software or computer program, called Expert System, which can Aid the Doctor in diagnosing a Disease. This expert system will surely bring down the mistakes which in turn will be boon for humanity. One such system to diagnose various kinds of diabetes, using the concept of symptoms, ANN and k-NN on MATLAB R2009a platform, has been developed in this work. In the present work a prototype expert system has been developed for clinical diagnostics by training the Artificial Neural Network and k Nearest Neighbor for Diabetes Mellitus diagnosis. The design aligns the structure and work of diagnostics mechanism similarly to the decision making process of a physician by dividing the diagnostics in three parts. The first part of the diagnostics is to find the risk of patient suffering from diabetes using weighted symptoms of the patient. Second part is to find output class value of diabetes of a patient using ANN (Artificial Neural Network) and k-NN (closer the output value to 1 more the patient is diabetic sufferer). Third part of our Project classifies whether the patient is Type 1 or Type 2 Diabetic along with respective Diet charts and preventive prescriptions. Standalone .exe application of the above project has been made to serve as potable project that can run on any Windows 95 or later