Introduction: Among sleep-related disorders, Sleep apnea has been under more attention and it’s the most common respiratory disorder in which respiration ceases frequently which can lead to serious health disorders and even mortality. Polysomnography is the standard method for diagnosing this disease at the moment which is costly and time-consuming. The aim of the present study was to analyze essential signals for the diagnosis of sleep apnea.Method: This analytical–descriptive was conducted on 50 patients (11 normal, 13 mild, 17 moderate and 9 severe patients) in the sleep clinic of Imam Khomeini hospital. Initially, data pre-processing was carried out in two steps(and Moving Average algorihtm). Next, using the SVD method, 12 features were extracted for airflow. Finally, to classify data, SVM with Quadratic, Polynomia and RBF kernels were trained and tested.Results: After applying different kernel functions on SVM, the RBF kernel showed the most efficient performance. After running the RBF kernel function ten times, the mean accuracy obtained for normal, apnea, and hypopnea modes were 92.74%, 91.70%, 93.26%.Conclusion: The results indicate that in online applications or applications in which volume and time calculations and the result are important simultaneously, patients could be diagnosed with acceptable accuracy using machine learning algorithms.

No abstract

Introduction: Sleep apnea syndrome can be considered as one of the most serious risk factors of sleep disorder. Due to the lack of information about this disease, many causes of unexpected deaths have been identified. With increasing the number of patients with this disease around the world, many patients suffer apnea complications.  Most of them are not treated because of the complex and costly and time-consuming polysomnography (PSG) diagnostic procedure.Material and Methods: This descriptive-analytical study was performed on 50 patients referred to sleep clinic of Imam Khomeini Hospital in Tehran,  Attempts to design, and develop a system for detection of sleep apnea and its severity using ECG signals, RR intervals and airflow. The random forest algorithm and MATLAB2016 were used in the design of the system that the algorithm inputs are extracted 8 features nonlinear in time-frequency domain from airflow and ECG signals and 10 nonlinear features of RR intervals.Results: The accuracy for normal, obstructive, central and mixed apnea was obtained at 95.3%, 97.92%, 99.60%, and 97.29%, respectively, and the accuracy For detection of normal, mild, moderate and severe apnea was obtained 96%, 94%, 94%, 96% respectively. According to the results, the proposed system can correctly classify the types of sleep apnea and its severity.Conclusion: The proposed system, which has high performance capability in addition to increasing the physician speed and accuracy in the diagnosis of apnea can be used in home systems and the areas where healthcare facilities are not sufficient.