Brain signal and image analysis using machine learning methods

• Main Author: Κανάς, Βασίλειος
• Other Authors: Σγάρμπας, Κυριάκος
• Format: Thesis
• Language: English
• Published: 2017
• Subjects: Brain signals; Machine learning; Speech recognition; Magnetic resonance imaging; Brain tumor segmentation; Synchronization; Regression; 616.804 754; Εγκεφαλικά σήματα; Μηχανική μάθηση; Αναγνώριση ομιλίας; Μαγνητική τομογραφία; Τμηματοποίηση εγκεφαλικού όγκου; Συγχρονισμός; Μαθηματική παλινδρόμηση

The work presented in this Ph.D manuscript focuses on implementing and designing machine learning for brain signal analysis. In the first part of the manuscript, dedicated to one-dimensional brain signal, we study electrocorticography (ECoG) signal processing for voice activity detection and syllable classification in order to design an interpretable and more efficient brain computer interface system for speech rehabilitation. The second part of this Ph.D dissertation is dedicated to two-dimensional biomedical signal analysis (image analysis). More specifically, we per-formed analysis of magnetic resonance medical images for brain tumor segmentation and grade classification. Finally, the last part of the thesis is based on mathematical modeling of biological neural networks. We aimed to study the microscopic dynamics of the brain neuronal networks through synchronization phenomena.