Data mining and mathematical modelling on diabetes

Malfunctioning of the β-cells of the pancreas leads to the metabolic disease known as diabetes mellitus (DM), which is characterized by significant glucose variation due to lack of insulin secretion, lack of insulin action, or both. DM can be broadly classified into two types: type 1 diabetes mellitus (T1DM) - which is caused mainly due to lack of insulin secretion and type 2 diabetes mellitus (T2DM) - which is caused due to lack of insulin action. The objectives of this work are to investigate on both types of DM to develop diagnostic and predictive models to simplify the clinical practice and the management of these chronic diseases. The insulin concentration and its influence on glucose one plays a key role in development of all types of DM disease. Since the 1960s, mathematical models of metabolism have been proposed in the literature. The majority of these models are glucose-based and have ignored the contribution of non-esterified fatty acid (NEFA) metabolism, which is an important source of energy for the body. Also, significant interactions exist among NEFA, glucose, and insulin. It is important to consider these metabolic interactions in order to characterize the endogenous energy production of a healthy or diabetic patient. The first objective of this thesis is to propose a new identification technique of a model of NEFA dynamics and was validated with experimental data obtained from oral glucose tolerance test (OGTT) in women with a history of gestational diabetes (GDM). GDM occurs when pregnant women without a previous diagnosis of diabetes develop a high blood glucose level. GDM could be a risk factor of development of type 2 DM. Since the publication of the Framingham algorithm for heart disease, tools that predict disease risk have been increasingly integrated into standards of practice. The utility of algorithms at the population level can serve several purposes in health care decision-making and planning. A risk prediction tool for Diabetes Mellitus (DM) can be particularly valuable for public health given the significant burden of diabetes and its projected increase in the coming years. The second objective of this work is to investigate the role of the shape of glucose curves in OGTT to define a model that is useful in clinical practice for prediction of the development of T2DM. The last part of the thesis is dedicated to T1DM. The most common intensive insulin treatment for T1DM requires administration of insulin subcutaneously 3 - 4 times daily in order to maintain normoglycemia (blood glucose concentration at 70 to 120 mg/dl ). Although the effectiveness of this technique is adequate, wide glucose fluctuations persist depending upon individual daily activity, such as meal intake, exercise, etc. The main challenge is to find a non-invasive continuous glucose monitoring system that permit providing comprehensive blood glucose profile without the need for numerous invasive finger-stick tests. Impedance spectroscopy could be a candidate for non-invasive continuous glucose monitoring in humans. The last objectives of this work is to study the relation between glucose and major electrolytes concentrations in blood to confirm that the impedance measurements will be adopted to monitor glucose concentration in blood