The doctoral dissertation in the field of Applied Physics will be examined at the Faculty of Science, Forestry and Technology, Kuopio campus.
What is the topic of your doctoral research? Why is it important to study the topic?
This doctoral research focuses on identifying optimal physiological signal combinations for automatic sleep staging, respiratory event detection, and obstructive sleep apnea (OSA) severity assessment. Polysomnography, the clinical gold standard for sleep assessment, requires multiple sensors to record brain activity, eye movements, airflow, respiratory effort, and blood oxygen saturation. Despite its diagnostic value, this complex measurement setup can cause patient discomfort, increase costs, and limit accessibility.
The research investigates whether deep learning methods can maintain reliable sleep assessment using reduced signal configurations. Identifying the most informative physiological signals could support the development of simplified, automated, and more accessible sleep monitoring systems, particularly for home-based assessment and OSA screening.
What are the key findings or observations of your doctoral research?
The findings demonstrate that accurate sleep assessment can be achieved using reduced physiological signal configurations, potentially decreasing the complexity of conventional sleep studies.
The results revealed negligible differences in electroencephalography (EEG) and electrooculography (EOG) signal characteristics between the left and right hemispheres. Furthermore, deep learning models achieved comparable sleep staging performance across different EEG and EOG configurations, indicating that reliable automatic sleep staging is feasible with fewer recording channels.
For respiratory event detection and obstructive sleep apnea (OSA) severity assessment, configurations including nasal pressure consistently achieved the highest performance, highlighting its importance for accurate respiratory monitoring. Although photoplethysmography (PPG) and blood oxygen saturation (SpO₂) showed lower accuracy in identifying individual respiratory events, their combination demonstrated potential for OSA screening.
Overall, these findings provide evidence supporting the development of simplified, automated sleep monitoring systems with reduced sensor requirements, potentially improving patient comfort, accessibility, and the feasibility of home-based sleep assessment.
What are the key research methods and materials used in your doctoral research?
This doctoral research comprised three studies utilizing polysomnography (PSG) recordings from independent datasets collected in Iceland, Australia, and the Netherlands.
Study I included 50 home-based PSG recordings collected in 2021 at the Reykjavik University Sleep Institute, Iceland. EEG and EOG signals were analyzed using amplitude and spectral analyses to investigate interhemispheric differences across sleep stages.
Study II utilized 876 in-laboratory PSG recordings collected between 2015 and 2017 at Princess Alexandra Hospital, Australia. Different EEG and EOG signal combinations were evaluated for automatic sleep staging using deep learning.
Study III included 604 in-laboratory PSG recordings collected in 2017 and 2019 at Kempenhaeghe, the Netherlands. Different cardiorespiratory signal combinations were evaluated using deep learning for respiratory event detection and OSA severity assessment.
The doctoral dissertation of Masoumeh Tashakori, MSc, entitled Optimal signal combinations for automatic sleep assessment and respiratory event detection will be examined at the Faculty of Science, Forestry and Technology, Kuopio campus. The opponent will be Associate Professor Kaare Mikkelsen, Aarhus University, Denmark, and the custos will be Professor Timo Leppänen, University of Eastern Finland. Language of the public defence is English.
