Estimation and classification of sleep apnea in adults by developed preprocessing?neural network models

Sleep Apnea Syndrome (SAS) is one of the serious worldwide health problems. Early diagnosis is an important factor in the treatment of the syndrome.In this study, various methods for estimating and classifying SAS were investigated. The first method, concerns estimation of SAS by an Artificial Neural Network (ANN) designed to employ the energy of particular scheme emerged through time?frequency analysis of snoring signals that linked to apnea. The second method, concerns the EEG signals taken from patients. The characteristic features, such as the Quadratic Phase Couplings (QPCs), exhibited by bi?spectrum of delta, theta, alfa, beta and gamma sub-bands of EEG were quantified and fed to ANN. The third method, concerns thoracic and abdominal signals taken from patients. These data were then split into wavelet coefficients up to 7th level through Discrete Wavelet Transform (DWT). A particular DWT?NN for classification of SAS was designed. The energies of coefficients of each detail (1?7 level) and the 7th approximation level were fed to the input of the ANN. The data evaluated through ANN were also evaluated through a specifically designed Adaptive Neuro?Fuzzy Inference System (ANFIS) and the obtained results were cross?compared.With the proposed methods, SAS was estimated and classified with highly significant success rates. Such data analysis may also be used in neurology and sleep disorder fields. The developed Wavelet?NN or DWT?NN model may be considered to integrate into the PSG system to provide ease both to medical specialists and patients.