Studies on diabetes analysis using human tongue images

Abstract

Studies on diabetes analysis using human tongue images newlineDiabetes is a metabolic disorder characterized by abnormal newlinepresence of glucose in the blood due to insufficient insulin secretion or the newlinepancreas is resistant to the produced insulin. Non-invasive and continuous newlinemonitoring of glucose concentration in blood is one of the most challenging newlineand exciting applications in the present day world. The traditional invasive newlinemethods available to detect diabetes are Fasting Plasma Glucose (FPG), newlineRandom Plasma Glucose (RPG), Oral Glucose Tolerance Test (OGTT) etc. newlineNon-invasive methods include Diabspot, association rule mining, K-Nearest newlineNeighbour (KNN), Artificial Neural Networks (ANN), B-Colouring newlinetechniques etc. Inspection of the human tongue images provides a simple, newlineimmediate, inexpensive, non invasive, early warning signal which conveys newlineenormous information about the internal organs of the human body. Initially, newlinethe whole tongue is segmented and gradually the experiments reveal that newlinesegmenting papillae region of the tongue contain enormous information with newlinerespect to taste buds, pancreas which ultimately paves way for diabetes newlinedetection in its earlier stages.In this thesis papillae of the tongue image is acquired using a USBMicroscope Camera (UMC), which has a pixel resolution of 1280x960, and newlineproduces 45FPS. It comprises of USB 3.2 connectivity. Through UMC, newlinetexture accuracy and magnification of 10 to 230 times can be achieved. The newlinehidden details in the object are captured via the coaxial lighting present in the newlineUMC camera. The digitized tongue images undergo many preprocessing steps newlinesuch as inspecting the tongue image, resizing it, smoothing a noisy tongue newlineimage, to suppress the unwanted distortions by background subtraction and newlineenhances papillae tongue image features for further processing. Earlier newlineresearchers have focused on segmenting the tongue based on color variations newlineof the tongue and the geometrical parameters of the tongue such as its length, newlinewidth, area, center distance, square area etc. The research work i