Skip to Main content Skip to Navigation

Analyse de l'évolution de la glycémie des patients diabétiques insulinodépendants

Abstract : The main goal of this thesis is to help type 1 diabetes (T1 D) contrai and stabilize blood sugar levels. Fo this, an analysis of the evolution of the blood glucose is necessary, then and after the recording of the blood glucose values using the CGM, a good method of predicting the glycemia is essential for the patient to adjust the blood sugar. dose of insulin injected on the basis of these predicted values. ln this context, we focused in the first chapter a study on the principle of the regulation of blood glucose, which we have presented glycemic homeostasis, the evolution of blood glucose, the organs responsible for the regulation of blood glucose, and sets of mechanisms for the regulation of blood glucose. Thus, to better understand diabetes, we presented generalities on diabetes: history of diabetes, distribution of diabetes in the world, types of diabetes and the difference between them, the means of treatment of type 1 diabetes and materials techniques used for the management of diabetes. ln the second chapter, we studied the evolution of blood glucose, so we showed that blood glucose has a chaotic appearance. As a result, blood glucose is unpredictable in the long term, with a predictability limit of almost 45 minutes. The third chapter was a continuation of the work presented in the preceding chapter. lndeed, after determining the predictability limit, we study the approaches to predict glucose levels. lndeed, a vast bibliographie researc has been launched on ail the methods of prediction of the glycemia of which one has the mathematical methods and the methods of artificial intelligence. ln this work, two approaches to predicting blood glucose have been proposed. The first approach is a new adaptive ANN. lndeed, by optimizing the ANN architecture f each patient. The precision of the proposed ANNs is discussed on the basis of certain statistical criteria such as RMSE and MAPE. The average obtained of RMSE is 6.43 mg / dl, and the average of MAPE is of 3.87% for a Horizon of Prediction HP= 15 min. Comparing with other technical models established in the literature, the proposed method has several advantages such as accuracy and adaptability. Thus, the experiments show the capacity of the proposed ANNs for a better prediction of the level of the glycemia. The second approach is a weighted SVR based on the DE algorithm, the average obtained from RMSE was 9.44 mg/ dl for an HP equal to 15 min. A comparison with the techniques established in the literature shows that the proposed method has many advantages such as precision, adaptability and ease of application. Base on the experimental results, the proposed combination of the SVR optimization algorithm with DE has im roved rediction accurac due to its efficienc in modelin nonlinear and corn lex data sets.
Document type :
Complete list of metadatas

Cited literature [203 references]  Display  Hide  Download
Contributor : Abes Star :  Contact
Submitted on : Friday, February 28, 2020 - 7:27:08 PM
Last modification on : Saturday, February 29, 2020 - 1:35:40 AM
Long-term archiving on: : Friday, May 29, 2020 - 6:23:15 PM


Version validated by the jury (STAR)


  • HAL Id : tel-02494566, version 1


Takoua Hamdi. Analyse de l'évolution de la glycémie des patients diabétiques insulinodépendants. Ordinateur et société [cs.CY]. Université de Toulon; Ecole Nationale des Sciences Informatiques (Tunis), 2019. Français. ⟨NNT : 2019TOUL0004⟩. ⟨tel-02494566⟩



Record views


Files downloads