Developing an sleep scorer by using Biosignals in Matlab. : Evaluation for sleep apnea patients.

Abstract: Nowadays, sleep disorders e.g. sleep apnea —the cessation of airflow at the nose and mouth lasting at least 10 second— are a broadly problem around the world. Direct and indirect costs associated to sleep problems are outsize and the quality of patient life is deteriorated because of it. In addition, Sleep is a fundamental part of everyday life, the lack of it or the poor quality of sleep may lead into the development of important diseases. Sleep studies are usually carried out by specialists by means of polysomnography. Polysomnography is a type of sleep study which is consisting of EEG, EOG, EMG, ECG, respiratory signals and/or many other biosignals which together can be used to determine the state of patient’s sleep and any other issue. Nowadays, visual inspection of these signals forms the “gold standard” in sleep clinics. The cost of monitoring a person overnight, the scarcity of beds available and the uncertainty of whether the results are representative of a normal nights’ sleep means that a move to home diagnostics is likely to be advantageous. Therefore, a necessity for home recorders systems capable of perform this kind of analysis has come out. A state machine based automatic scorer is developed and evaluated in Matlab by using 12 recordings of apnoeic patients from sleep heart health study (SHHS) database. By the analysis of EEG, EOG, EMG, Oxygen saturation (Sao2) and respiratory movements signals, the implemented algorithm is trained and evaluated to detect the five stages of subject’s sleep (Wake, N1, N2, N3, or REM) as well as apnoeic episodes according to guidelines from American Academy of Sleep Medicine (AASM). In the final evaluation of algorithms, the automatic scorer achieved 74±5.27% accuracy for all five stages and Cohen’s kappa of 0.5 for the overall set of 12 patients, being the accuracy better for healthier subjects and reaching in this case 78±4.05%. The analysis of the sleep apnea concluded with a sensitivity of 47.08%, a specificity of 83.38%, and an accuracy of 78.1%. Differences in the performance among patients according to their apnea/hypopnea index were significant.   Key Words: Polysomnography, AASM, Sleep apnea/hypopnea.