چكيده به لاتين
One of the most common sleep disorders is sleep apnea. Sleep apnea is a relatively common condition caused by sleep apnea. Research has shown that by controlling and preventing respiratory arrest, the complications of this disease can be prevented. At present, the main standard for diagnosing sleep apnea is polysomnography. This device provides important information by recording various activities such as electroencephalogram, electroacogram, electromyography, electrocardiogram, oximetry, air flow and respiratory activities, and can be used to assess the severity of the disease. Of course, this diagnostic test is very expensive and time consuming. In addition, access to sleep clinics is not easy, diagnosis of the disease using simpler signals can be very valuable because they are easier to register and without the need for special places to register (e.g. Sleep clinics) is possible. Recently, many attempts have been made by researchers to diagnose the disease using fewer signals than PSG. The aim of this study was to automatically detect and classify obstructive sleep apnea (OSA) based on the application of classification algorithms. In this study, two new methods for automatic diagnosis and classification of apnea and non-apnea events were presented. The Apnea-ECG physiotherapy database, which contains 70 registrations of single-lead ECG signals, was used to evaluate the proposed methods. In the first method, in the preprocessing section, an automatic method for detecting and removing noise windows was implemented. The proposed feature is extracted based on the experimental state analysis (EMD) algorithm. In addition, the extracted features are classified by different machine learning methods. Then, evolutionary algorithms have been used to improve the performance of MLPNN in OSA detection, with the Particle Swarm Optimization (PSO) algorithm having the highest performance among the other methods used. In the second method, in the preprocessing stage, the ECG signal is first divided into 2-second sections, then the RP algorithm is used to convert the signal into two-dimensional images. Finally, transmission learning for adaptation of pre-trained deep conduction neural networks (DCNNs) has been used to find the most appropriate method for classifying obstructive sleep apnea (OSA) using an electrocardiogram (ECG) signal.
