Automated Analysis of the Oximetry Signal to Simplify the Diagnosis of Pediatric Sleep Apnea: From Feature-Engineering to Deep-Learning Approaches (Springer Theses)

Supervisors’ Foreword\nForeword by the Bioengineering Group of the Automatic Control Spanish Association (CEA)\nAcknowledgements\nContents\nAcronyms\n1 Introduction\n	1.1 The Biomedical Signal Processing Framework: Feature-Engineering and Deep Learning\n	1.2 Pediatric Obstructive Sleep Apnea (OSA)\n	1.3 Pediatric OSA Diagnosis: Polysomnography (PSG)\n	1.4 Alternatives to PSG\n		1.4.1 Overnight Oximetry\n	1.5 State-of-the-art: Automated Analysis of the Oximetry Signal to Diagnose Pediatric OSA\n		1.5.1 Conventional Oximetric Indices\n		1.5.2 Automated Signal Processing Methods\n	References\n2 Hypotheses and Objectives\n	2.1 Hypotheses\n	2.2 Objectives\n	2.3 Thesis Organization\n	References\n3 Methods\n	3.1 Subjects and Signals Under Study\n		3.1.1 Childhood Adenotonsillectomy Trial (CHAT) Database\n		3.1.2 University of Chicago (UofC) Database\n		3.1.3 Burgos University Hospital (BUH) Database\n	3.2 Pre-processing\n	3.3 Feature Engineering\n		3.3.1 Feature Extraction\n		3.3.2 Feature Selection\n		3.3.3 Pattern Recognition\n	3.4 Deep Learning\n		3.4.1 Proposed Convolutional Neural Network (CNN) Model\n		3.4.2 CNN Training and Optimization Process\n	3.5 Statistical Analysis\n		3.5.1 Statistical Hypothesis Tests\n		3.5.2 Diagnostic Performance Metrics\n		3.5.3 Measures of Agreement\n		3.5.4 Validation Strategies\n	References\n4 Results\n	4.1 Application of Novel Feature-Extraction Algorithms\n		4.1.1 Bispectral Analysis\n		4.1.2 Wavelet Analysis\n		4.1.3 Detrended Fluctuation Analysis\n	4.2 Application of Deep-Learning Techniques\n	References\n5 Discussion\n	5.1 Novel Features to Provide Relevant and Complementary Information from Oximetry Recordings\n		5.1.1 Bispectral Analysis\n		5.1.2 Wavelet Analysis\n		5.1.3 Detrended Fluctuation Analysis\n	5.2 A Deep-Learning Based Methodology to Automatically Extract the Relevant Information from Raw Oximetry Recordings\n	5.3 Comparison of Performance: Feature-Engineering,  Deep-Learning, and State-of-the-art\n		5.3.1 Comparison Between Feature-Engineering and Deep-Learning Approaches\n		5.3.2 Comparison with State-of-the-art Studies\n	5.4 Limitations of the Study\n	References\n6 Conclusions\n	6.1 Contributions\n	6.2 Main Conclusions\n	6.3 Future Research Lines\n	References\nAppendix A About the Author