Essay: Existing methodologies and techniques for early diagnosis of diabetes mellitus

Healthcare is such a wide and extensive area of research wherein diabetes is such a deadly disease which hampers a common man life at the extreme end. Under healthcare, there is always a chance for uncertainty and imprecision under various aspects of medical diagnosis process. This paper reviews various existing methodologies and techniques used for early diagnosis of diabetes mellitus which works like risk alarms to save human life. This survey gives us current state of research in diagnosing disease like diabetes and helps us to find out difficulties with the existing systems. In many of the expert systems different soft computing and data mining techniques are used along with usage of real time dataset from hospitals or readily available datasets like PIMA indian dataset, colic dataset to diagnose the disease.
I.Introduction
Diabetes is a metabolic disorder in which body is unable to handle sugar characterized by hyperglycemia i.e. high blood glucose levels resulting from defects in insulin secretion. It is a silent killer. The chronic hyperglycemia of diabetes is associated with disturbances of carbohydrate, fat and protein metabolism which in return results in long-term damage, dysfunction and failure of different organs, especially the eyes, kidneys, nerves, heart and blood vessels. The body needs insulin to use sugar, fat and protein from the diet for day to day activities [1]. In spite of so much medical progress still many of diabetes mellitus are unaware of their disease as many of the symptoms are common with other diseases. So it takes long duration for diagnosis of such a diabetes disease.[23] Hence there is a need to develop such an expert system which not only alarms the risk of disease but also helps in finding out the solutions with expertise. In this review paper we focus on existing methods for diabetes detection so as to know the current developments in the field of diabetes under healthcare.
II. Commonly used approaches for diagnosis of diabetes:
A. Fuzzy approach
Nitin Bhatia and sangeet kumar proposed an FCM (fuzzy cognitive maps) approach to model knowledge-based systems for diagnosing thyroid diagnosis. They used temporal medical data. Proposed framework adaptive algorithm used for learning FCM which works at three levels to record symptoms. Software tool is used under soft computing which tests for 50 cases and got accuracy of 96%.[3], [20]
Mythili Thirugnsnsm proposed to improve diabetes diagnosis using three combined approaches of fuzzy, Neural network and case based reasoning. All three are dependent to each other and which gives accuracy in prediction rate for occurrence of diabetes mellitus.[1]
Chang-Shing Lee defined a fuzzy diabetic ontology (FDO) to model the diabetes knowledge. Instances of FDO are generated by the fuzzy ontology generating mechanism. It simulates fuzzy decision making applications.[14],[18]
Paul Grant proposed an approach for controlling of diabetes through fuzzy logic and insulin pump which is small wearable device for short acting insulin. When sensor senses the deficiency then optimal insulin infusion rate has been calculated and through fuzzy logic patient has been operated through a pump. It avoids multiple daily injections.[19]
Along with fuzzy logic faith-Michael Emeka Uzoka, Okure Obot, Ken Barker introduced AHP which is analytical hierarchy process. It discovered from the fuzzy logic diagnosis covary a little bit more strongly to the conventional diagnosis results than AHP. Fuzzification and inference engine is used. [22]
Microalbuminuria is predictor of diabetes and other diseases like cardiovascular. Hamid Marateb, Marjan, Elham Faghihimani, Masoud , Dario Farina proposed expert based fuzzy MA classifier in which with rule induction particle sworm optimization is performed. After classifying variety of classifiers based on different diabetic parameters like BMI, HBA1c, age etc. performance is calculated on 10 folds cross validation. Hence parameters accessed are Sensitivity, Specificity, precision and accuracy for values 95 %,85%,84% & 92% resp. [24]