Özet:

Recently, one of the most important problems in society, sleep disorders have seriously affected the health and quality of life of individuals. There are disorders caused by many sleep disorders, such as insomnia, narcolepsia, sleep apnea, and unhappy leg syndrome. The main factor that causes sleep disorders is sleep disorders that can not result in a person's instant awakening, which decreases the quality of sleep. These interruptions are temporarily awakening with another name and represent a sudden change in the pattern of a brain wave (Electroencephalogram -EEG) activity. Aerousal detection is usually done using EEG data according to the criteria determined by the American Academy of Sleep Medicine (AASM). The aim of this study is to detect arousal in patients through EEG signals according to the criteria determined by AASM. For this purpose, first and foremost, the EEG signals of the single channel (C3/A2) of the 5 patients involved in the study were filtering, normalization and segmentation predictions, respectively. Through the characterization process that was subsequently carried out with the methods of Spectral Strength (Power Spectral Density-PSD) and Discrete Wavelet Transform-DWT (Discrete Wavelet Transform-DWT), the EEG signal segments have created 2 characteristics sets and the combination of these characteristics sets, the 3 characteristics sets. Three properties were then identified using the Wrapper Subset Evaluation (WSE) method to select the properties. EEG segments that contain arousal have been identified by the classification of the characteristics defined by the Artificial Neural Networks (YSA) and the Algorithms of the Emergency Forest (RO). These studies have found that the EEG signals were very successful with only single-channel EEG signals, without any other PSG signals recording needed. As a result of the study, the effective features of the Specialty Set 3 and the highest accuracy rate with the YSA were achieved at 99.5%.

Anahtar Kelimeler:

Özet:

Nowadays, sleep disorders, one of the most important problems of society, affect the health and quality of life of individuals seriously. There are diseases caused by many sleep disorders such as insomnia, narcolepsy, sleep apnea and restless legs syndrome. The main factor that causes sleep disorders is sleep interruptions that don’t result in the awakening of the individual during sleep and reduce sleep quality. Arousal, also known as awakening, are these temporary interruptions and represent a sudden change in the pattern of brainwave (Electroencephalogram-EEG) activity. Generally, Arousal detection is performed by utilizing EEG datas according to the criterions determined by American Academy of Sleep Medicine (AASM). This study's purpose is to detect arousals of patients through EEG, in accordance with the criterions of ASSM. For this purpose, primarily, filtering, normalization and segmentation pre-processes were applied to single channel (C3 / A2) EEG signals of 5 patients included in the study, respectively. Afterwards 2 feature sets and 3rd feature set by combining these 2 feature sets were generated for EEG signal segments thanks to feature extraction processing carried out Power Spectral Density (PSD) and Discrete Wavelet Transform (DWT) methods. Subsquently, efficient features were determined by implementing Wrapper Subset Evaluation (WSE) that is feature selection method on to the 3 features sets created. Finally, the EEG segments containing arousals were detected by classifying the determined features by Artificial Neural Networks (ANN) and Random Forest (RO) algorithms. With these studies performed , it has been determined to achieve successful results with only single-channel EEG signals without the need for any PSG signal recording other than EEG signal recording. As a result of the study, the highest accuracy rate (99.05%) were obtained by using the effective features of Feature Set 3 and ANN.