Detection of ECG ST segment changes using time frequency transforms for the early diagnosis of myocardial infarction

Early detecetion of ST segment's depression or elevation is very important for prevention of myocardial ischemia and it is very important to prevent a myocardial infarction that may occur in the future. In the thesis study, methods based on time-frequency transforms have been developed in order to early detection of ST segment's depressions or elevations in the ECG R-R interval. In the first phase of the thesis study; a class was formed from RR intervals with healthy and arrhythmia and the other class was formed from RR intervals with ST segment changes. In the second phase; R-R intervals with arrhythmia, healthy, ST segment depression and ST segment elevation were selected to be four classes. The thesis study consist of database construction, feature extraction and classification. In the study large databases produced by selecting ECG R-R intervals from MIT-BIH Arrhythmia, European ST-T and Long Term ST databases were used. Smoothed Wigner-Ville, Choi-Williams, Born-Jordan, Bessel ve Zhao-Atlas-Mark time frequency transforms were used in the feature extraction. The classification performance results analyzed using Decision Tree, Support Vector Machines, K-nearest Neighbor and Ensemble Classifiers methods. The classification results are above the values belonging to the studies in the literatüre. In addition to the speed of the proposed algorithm is very suitable for telemedicine systems. The thesis study will be a part of the Decision Support System in the Telemedicine system being developed.