Lakhasly

Sleep studies support doctors in diagnosing sleep disorders such as insomnia, narcolepsy or sleep apnea and set the starting point for proper follow up. sleep stages are commonly assessed using the physiological criteria based on Electroencephalogram (EEG), Electro-oculogram (EOG), Electromyogram (EMG), heart rate,snoring, and respiration.Holland, Dement, and Raynal in 1974, proposed the term Polysomnography (PSG) for describing the recording, analysis and interpretation of multiple, simultaneous physiologic parameters recorded during sleep.This paper is organized as follows.

Sleep studies support doctors in diagnosing sleep disorders such as insomnia, narcolepsy or sleep apnea and set the starting point for proper follow up. sleep stages are commonly assessed using the physiological criteria based on Electroencephalogram (EEG), Electro-oculogram (EOG), Electromyogram (EMG), heart rate,snoring, and respiration. Holland, Dement, and Raynal in 1974, proposed the term Polysomnography (PSG) for describing the recording, analysis and interpretation of multiple, simultaneous physiologic parameters recorded during sleep. According to American Academy of Sleep Medicine (AASM) [1], sleep is divided in two main stages, Non-Rapid Eye Movement stage (NREM) which by itself includes three stages: N1, N2, and N3 and Rapid-Eye Movement stage (REM). Wakefulness is also included in analysis as the fifth stage. In the manual process of sleep scoring each 30 s of PSG signals, called an epoch, is associated with one of five sleep stages. This process is tedious, subjective and costly.Therefore, over the last years, extensive research efforts have been taken to purse for an automatic sleep stage classification based on machine learning techniques. Most of the existing work rely on the hand-crafted features that require prior knowledge of the domain expert. Additionally, these systems are strongly dependent to available labeled data. Recently, deep learning method have been used in sleep stage classification systems, trying to enhance the performance of feature extraction and classification steps.Inspired by the ability of deep networks, this work proposes a method for classification of spectrogram images, extracted from each epoch of sleep EEG, using a pre-trained Convolutional Neural Network, called AlexNet. This paper is organized as follows. Section 2 gives a brief overview of how sleep stage classification using deep learning methods is treated in the literature. Section 3 provides detailed description of the proposed algorithm. Section 4, the process of performance evaluation is described. Section 5, the evaluation results are presented and discussed. Section 6 finalizes the paper with conclusions and future directions.