Analysis of GT dispersion on obstructive sleep apnea syndrome patients

Sleep apnea is a sleep disorder characterized by pauses in breathing during sleep. The standard definition of any apneic event includes a minimum 10 second interval between breaths. In obstructive sleep apnea (OSA), breathing is interrupted by a physical block to airflow despite respiratory effort. The interrupts in breathing have been shown to affect the functioning of the heart, causing significant changes on electrocardiographic (ECG) wave shapes. The main purpose of this thesis is to develop a robust algorithm to automatically determine the fiducials on each heart beat, such as Q, R, S, and T-onset and T-end, on long sleep ECG recordings. Our second aim was to test our algorithm on different ECG wave shapes and noise types. The final aim of this thesis was to apply the algorithm on sleep ECG recordings coming from OSA patients and a control group (non-OSA), and to statistically analyze the effect of OSA and sleep stages on ECG intervals such as QT and RR.In this thesis, a Matlab-based automatic fiducial detection algorithm using two approaches was developed: (1) A simple derivative-based algorithm to determine the Q, R, and S points on each heart beat. (2) A Haar wavelet transform-based algorithm to determine the onset and ending points of the T-wave on each heart beat. In order to test our algorithm and perform the analysis on the effects of OSA and sleep stages on the QT and RR intervals, 25 sleep ECG recordings (14 OSA syndrome subjects, 11 non-OSA syndrome subjects) were used which were previously obtained during sleep studies performed in Ankara Gülhane Military Medical Academy (GATA) psychiatry clinic sleep laboratories. First, a sample dataset was created that included representative heart beats possessing different QRS and T-wave shapes to test the performance of our automatic algorithm. On this dataset two cardiologists from Kahramanmaraş State Hospital annotated Q, and T-end fiducials using a Matlab-based graphical user interface. R points were annoted by the author of the thesis. The relative error between the corrected QT (QTc) and RR intervals determined by the cardiologists and the results from the algorithm were computed, and found that QTc and RR intervals could be determined with a relative error of 4.47% and 1.31%, respectively.Finally, automatically detected QTc, QTc dispersion (the difference between the minimum and maximum QTc values on each heart beat, QTcd) and RR intervals for different sleep stages were analyzed using t-test on OSA syndrome and non-OSA syndrome subjects, independently. Our analysis also included a statistical comparison of the same algorithm-determined-parameters between OSA syndrome and non-OSA syndrome subjects. As a result of these tests, we found that there was a significant difference in QTc interval between OSA syndrome and non-OSA syndrome populations for different sleep stages.KEY WORDS: Obstructive sleep apnea, QT dispersion, wavelet transform, t-testAdvisor: Assist. Prof. Dr. Bülent Yılmaz, Başkent University, Department of Biomedical Engineering