Analysis of EEG transient waveforms in sleep disorders

Transient waveforms (k-complexes, sleep spindles, arousal, etc.) happens instantaneously in sleep, have distinctive structural features, amplitudes and frequencies, and are difficult to distinguish from the background of electroencephalography (EEG) which are called the microstructure of the EEG. The analysis of these waveforms is important for brain research, sleep studies, sleep stage scorings and assessment of sleep disorders. In literature, related works are used the specified feature extraction of the transient waveforms and classification studies. In this study, different from the literature, it is aimed to develop several methods to detect high accuracy three different waveforms with high accuracy, unlike single transient EEG waveform. In this study, the EEG recordings were obtained from the National Sleep Research Resource, DREAMS databases and Ankara Gulhane Military Medical Academy, Psychiatry Department the Sleep Laboratory. In the first part of the study, signal processing methods were used to analyze patients' EEG signal records. Discrete Fourier transform, power spectrum methods in spectral analysis; short - time Fourier transform in time-frequency analysis; discrete and continuous wavelet transform in the wavelet analysis were used. In wave form analyzes, signal parameters such as effective value of EEG energy, sub-band energy analysis, cross correlation function, zero cross rate, mean and variance were investigated. According to results, cross correlation function, sub-band energy analysis and zero cross rate for k-complex; the sum of the continuous wavelet transform coefficients in the frequency range for sleep spindles; and the mean and variance of the continuous wavelet transform coefficients for arousal were determined as features. In the second part of the study, with determined features, transient waveforms were classified as "existence" or "absence" using linear discriminant analysis, support vector machines and artificial neural networks. As a result of the study, a decision support system algorithm, can detect transient waveforms successfully, was developed that can yield results better than or as equivalent as the literature.