| Titre : | A new AI-IoT system for the Noninvasive glucose monitoring |
| Auteurs : | OTHMANE AMANI, Auteur ; Samir Bourekkache, Directeur de thèse ; Imane YOUKANA, Directeur de thèse ; Laïd Kahloul, Directeur de thèse |
| Editeur : | Biskra [Algérie] : Faculté des Sciences Exactes et des Sciences de la Nature et de la Vie, Université Mohamed Khider, 2024 |
| Format : | 1 vol. (54 p.) / ill., couv. ill. en coul / 30 cm |
| Langues: | Anglais |
| Mots-clés: | Diabetes, Non invasive, Artificial intelligence, Blood Glucose Level, Photoplethysmography (PPG), Galvanic skin response (GSR), IoT. |
| Résumé : |
Diabetes mellitus is a global health concern with rising complications. Traditional blood sugar monitoring methods rely on invasive techniques requiring finger pricking, which can be inconvenient and discourage frequent checks. on the other hand, noninvasive blood glucose monitoring offers a promising alternative, potentially improving patient compliance and diabetes management. This study propose to develop a non-invasive technology for measuring blood glucose levels using the internet of things (IoT) devices: photo-plethysmography (PPG) and galvanic skin response (GSR) sensors together, and investigating the efficacy of various machine learning and deep learning regression algorithms. We used PPG and GSR sensors to collect data from 50 individuals at the Hakim Saadan Hospital in Biskra, ensuring ethical and controlled data collection. The proposed system integrates a comfortable wearable device and a secure cloud platform. It empowers patients with diabetes to gain real-time insights into their glucose levels, allowing for proactive self-care. In addition, in order to facilitate the data visualization, a Django web application is developed. This web platform can serve as a valuable tool for managing and exploring the collected data. The research demonstrates that combining PPG and GSR data yields superior results compared to PPG alone. Among the investigated AI models, the Long Short-Term Memory (LSTM) model achieved the highest accuracy in blood sugar prediction. These findings contribute significantly to the development of non-invasive blood glucose monitoring techniques, particularly with regards to leveraging multi-modal data fusion and improving patient compliance. |
| Sommaire : |
Acknowledgements ii Dedication iii Conferences and Animations v Accepted Papers vi Abstract vii Résumé viii 1 General Introduction 1 1.1 Context . . . . . . . . . . . . . . . . . . 1 1.2 Problematic and Motivation . . . . . . . . . 2 1.3 The work’s purpose . . . . . . . . . . . . . 2 1.4 Structure of the dissertation . . . . . . . . 2 2 Background 4 2.1 Introduction . .. . 4 2.2 Healthcare . . . . . . . . . . . .. 4 2.2.1 Types of health care . . .. . . . 4 2.3 Diabetes . . . . . . . . . . . . 5 2.3.1 Types of diabetes . . . . . . . . . . . 6 2.3.2 Symptoms and Complications . . . . . . . . . . . . . . . 6 2.3.3 Importance of blood glucose monitoring . . . . . . . . 7 2.4 Artificial Intelligence . . . . . . . . . . . 7 2.4.1 Machine learning . . . . . . . . . . . . . . 8 2.4.1.1 Types of Learning . . . . . . .. . . 8 2.4.2 Deep learning . . . . . . . . . . . . . . . . . . . . 11 2.4.3 The differences between Machine Learning and Deep Learning . . 12 2.4.4 AI in Healthcare . . . . . . .. . . . . . . . 12 2.4.4.1 Some application of AI in healthcare . . . . . . 13 2.5 Internet of Things . . . . . . . . . .. . 13 2.5.1 IoT architecture . . . . . . . .. 13 2.5.2 IoT in healthcare . . . . . . . .. 16 2.6 An overview of Blood Glucose monitoring techniques . . . . . 18 2.6.1 Optical Methods . . . . . . . . . . . . . . . . . . . 19 2.6.1.1 Types of Optical Methods . . . . . . 19 2.6.1.2 Photoplethysmography (PPG) . . . . . . . 19 2.6.1.3 Relation between Photoplethysmography and Heart Rate 20 2.6.1.4 Relation between Photoplethysmography and Oxygen saturation. . . . . . 21 2.6.1.5 Relation between Photoplethysmography and glucose levels 22 2.6.1.6 Related work about non-invasive blood glucose monitoring using Photoplethysmography (P. 23 2.6.2 Physiological Methods . . . . . . . . . . . . . .. 23 2.6.2.1 Galvanic Skin Response (GSR) . . . . . . 24 2.6.2.2 Relation between Galvanic Skin Response and glucose levels 24 2.6.2.3 Related work about non-invasive blood glucose monitoringusing GSR . . . 24 2.6.2.4 Related work about non-invasive blood glucose monitoring combining PPG and GSR . 25 2.6.3 Motivation for proposed Non invasive glucose monitoring . . . . . 25 2.6.4 Our contributions . . . . . . . . . 27 2.7 Conclusion . . . . . . . . . . . . .. . . 28 3 Design and implementation 29 3.1 Introduction . . . . . . . . . . . . . . 29 3.2 Architecture of System . . . . . . 29 3.2.1 Hardware part . . . . . . . . . . . . . . 30 3.2.1.1 Hardware architecture . . . . . . 32 3.2.2 Software part . . . . . . . . . . . . . . . . 33 3.2.2.1 Datasets Collection . . . . . . . 33 3.2.2.2 Data Real-Time Storage . . . . . . 3.2.2.3 Data Preparation . . . . . . . . . . . . . . . . 34 3.2.2.4 Blood glucose level prediction using machine learning model. . . . .35 3.2.2.5 Blood glucose level prediction using deep learning models 36 3.2.2.6 Performance Evaluation Metrics . . . . . . . . 39 3.2.3 The 3D design . . . . . . . . . . . . . . 40 3.3 Diagrams . . . . . . . . . . .. . 41 3.3.1.1 Use case diagram . . . . . . . . . . . . . . 41 3.3.1.2 Sequence diagram . . . . . . . . 42 3.3.1.3 Class diagram . . . . . . . . . . 43 3.4 Implementation . . . . . . . . . . . . . . 43 3.4.1 Languages and tools for development . . . . . . . . . 43 3.4.1.1 Hardware tools . . . . . . . . . . . . . . 43 3.4.1.2 Software tools . . . . . . .. . . . 44 3.4.2 Hardware realisation . . . . . . . . . . . 46 3.4.3 Software realisation . . . . . . . . . . . . . 47 3.4.3.1 Sketch for Arduino . . . . . . . . . 47 3.4.3.2 Implementation of AI models . . . . . . . . . 5 |
| Type de document : | Mémoire master |
Disponibilité (1)
| Cote | Support | Localisation | Statut |
|---|---|---|---|
| MINF/887 | Mémoire master | bibliothèque sciences exactes | Consultable |
