Şebnem Yosunkaya

Efficient sleep stage recognition system based on EEG signal using k-means clustering based feature weighting

Sleep scoring is one of the most important methods for diagnosis in psychiatry and neurology. Sleep staging is a time consuming and difficult task conducted by sleep specialists. The purposes of this work are to automatic score the sleep stages and to help to sleep physicians on sleep stage scoring. In this work, a novel data preprocessing method called k-means clustering based feature weighting (KMCFW) has been proposed and combined with k-NN (k-nearest neighbor) and decision tree classifiers to classify the EEG (electroencephalogram) sleep into six sleep stages including awake, N-REM (non-rapid eye movement) stage 1, N-REM stage 2, N-REM stage 3, REM, and non-sleep (movement time). First of all, frequency domain features belonging to sleep EEG signal have been extracted using Welch spectral analysis method and composed 129 features from EEG signal relating each sleep stages. In order to decrease the features, the statistical features comprising minimum value, maximum value, standard deviation, and mean value have been used and then reduced from 129 to 4 features. In the second phase, the sleep stages dataset with four features has been weighted by means of k-means clustering based feature weighting. Finally, the weighted sleep stages have been automatically classified into six sleep stages using k-NN and C4.5 decision tree classifier. In the classification of sleep stages, the k values of 10, 20, 30, 40, 50, and 60 in k-NN classifier have been used and compared with each other. In the experimental results, while sleep stages has been classified with 55.88% success rate using k-NN classifier (for k value of 40), the weighted sleep stages with KMCFW has been recognized with 82.15% success rate k-NN classifier (for k value of 40). And also, we have investigated the relevance between sleep stages and frequency domain features belonging to EEG signal. These results have demonstrated that proposed weighting method have a considerable impact on automatic determining of sleep stages. This system could be used as an online system in the automatic scoring of sleep stages and helps to sleep physicians in the sleep scoring process.

Comparison of Different Classifier Algorithms on the Automated Detection of Obstructive Sleep Apnea Syndrome

In this paper, we have compared the classifier algorithms including C4.5 decision tree, le artificial neural network (ANN), artificial immune recognition system (AIRS), and adaptive neuro-fuzzy inference system (ANFIS) in the diagnosis of obstructive sleep apnea syndrome (OSAS), which is an important disease that affects both the right and the left cardiac ventricle. The goal of this study was to find the best classifier model on the diagnosis of OSAS. The clinical features were obtained from Polysomnography device as a diagnostic tool for obstructive sleep apnea in patients clinically suspected of suffering this disease in this study. The clinical features are arousals index, apnea–hypopnea index (AHI), SaO2 minimum value in stage of rapid eye movement, and percent sleep time in stage of SaO2 intervals bigger than 89%. In our experiments, a total of 83 patients (58 with a positive OSAS (AHI > 5) and 25 with a negative OSAS such that normal subjects) were examined. The decision support systems can help to physicians in the diagnosing of any disorder or disease using clues obtained from signal or images taken from subject having any disorder. In order to compare the used classifier algorithms, the mean square error, classification accuracy, area under the receiver operating characteristics curve (AUC), and sensitivity and specificity analysis have been used. The obtained AUC values of C4.5 decision tree, ANN, AIRS, and ANFIS classifiers are 0.971, 0.96, 0.96, and 0.922, respectively. These results have shown that the best classifier system is C4.5 decision tree classifier on the diagnosis of obstructive sleep apnea syndrome.

Sleep spindles recognition system based on time and frequency domain features

Sleep spindle is the one of important components determining N-REM (Non-Rapid Eye Movement) stage 2 in the sleep stages. The symptoms of N-REM stage 2 are sleep spindle and K-complex and here sleep spindles are automatically recognized by using time and frequency domain features belonging to EEG (Electroencephalograph) signals obtained from three patient subjects. In this study, the proposed method consists of two steps. In the first step, six time domain features have been extracted from raw EEG signals. As for the extraction of frequency domain features from raw EEG signals, Welch spectral analysis has been used and applied to raw EEG signals. By this way, 65 frequency domain features have been extracted and reduced from 65 to 4 features by using statistical measures including minimum, maximum, standard deviation, and mean values. Three feature sets including only time domain, only frequency domain, and both time and frequency domain features have been used and the numbers of these feature sets are 6, 4, and 10, respectively. In the second step, artificial neural network (ANN) with LM (Levenberg-Marquardt) has been used to classify the sleep spindles evaluated beforehand by sleep expert physicians. The obtained classification accuracies for three features sets in the classification of sleep spindles are 100%, 56.86%, and 93.84% by using LM-ANN (for ten node in hidden layer). The obtained results have presented that the proposed recognition system could be confidently used in the automatic classification of sleep spindles.

Multi-class f-score feature selection approach to classification of obstructive sleep apnea syndrome

In this paper, a new feature selection named as multi-class f-score feature selection is proposed for sleep apnea classification having different disorder degrees (mild OSAS, moderate OSAS, serious OSAS, and non-OSAS). f-Score is used to measure the discriminating power of features in the classification of two-class pattern recognition problems. In order to apply the f-score feature selection to multi-class datasets, we have used the f-score feature selection as pairwise (in the form of two classes) in the diagnosis of obstructive sleep apnea syndrome (OSAS) with four classes. After feature selection process, MLPANN (Multi-layer perceptron artificial neural network) classifier is used to diagnose the OSAS having different disorder degrees. While MLPANN obtained 63.41% classification accuracy on the diagnosis of OSAS, the combination of MLPANN and multi-class f-score feature selection achieved 84.14% classification accuracy using 50-50% training-testing split of OSAS dataset with four classes. These results demonstrate that the proposed multi-class f-score feature selection method is effective and robust in determining the disorder degrees of OSAS.

Pairwise ANFIS Approach to Determining the Disorder Degree of Obstructive Sleep Apnea Syndrome

Obstructive sleep apnea syndrome (OSAS) is an important disease that affects both the right and the left cardiac ventricle. This paper presents a novel classification method called pairwise ANFIS based on Adaptive Neuro-Fuzzy Inference System (ANFIS) and one against all method for detecting the obstructive sleep apnea syndrome. In order to extract the features related with OSAS, we have used the clinical features obtained from Polysomnography device as a diagnostic tool for obstructive sleep apnea (OSA) in patients clinically suspected of suffering from this disease. The clinical features obtained from Polysomnography Reports are Arousals Index (ARI), Apnea and Hypoapnea Index (AHI), SaO2 minimum value in stage of REM, and Percent Sleep Time (PST) in stage of SaO2 intervals bigger than 89%. Since ANFIS has output with one class, we have extended the output of ANFIS to multi class by means of one against all method to diagnose the OSAS that has four classes consisting of normal (25 subjects), mild OSAS (AHI = 5–15 and 14 subjects), middle OSAS (AHI < 15–30 and 18 subjects), and heavy OSAS (AHI > 30 and 26 subjects). The classification accuracy, sensitivity and specifity analysis, mean square error, and confusion matrix have been used to test the performance of proposed method. The obtained classification accuracies are 82.92%, 82.92%, 85.36%, and 87.80% for each class including normal, mild OSAS, middle OSAS, and heavy OSAS using ANFIS with one against all method with 50–50% train-test split, respectively. Combining ANFIS and one against all method that is firstly proposed by us was firstly applied for diagnosing the OSAS. The proposed method has produced very promising results in the detecting the obstructive sleep apnea syndrome.

Automatic sleep staging based on SVD, VMD, HHT and morphological features of single-lead ECG signal

Abstract Electroencephalogram (EEG) signals, which are among the primary polysomnography (PSG) signals used for sleep staging, are difficult to obtain and interpret. However, it is much easier to obtain and interpret electrocardiogram (ECG) signals. The use of ECG signals for automatic sleep staging systems could bring practicality to these systems. In this study, ECG signals were used to identify the wake (W), non-rapid eye movement (NREM) and rapid eye movement (REM) stages of the sleep data from two different databases with 17,758 epochs of 28 subjects (21 healthy subjects and 7 obstructive sleep apnea (OSA) patients) in total. Four different methods were used to extract features from these signals: Singular Value Decomposition (SVD), Variational Mode Decomposition (VMD), Hilbert Huang Transform (HHT), and Morphological method. As a result of applying the methods separately, four different data sets were obtained. The four different datasets were given to the Wrapper Subset Evaluation system with the best-first search algorithm. After the feature selection procedure, the datasets were separately classified by using the Random Forest classifier. The results were interpreted by using the essential statistical criteria. Among the methods, morphological method was the most successful and it was followed by SVD in terms of success rate for both two databases. For the first database, the mean classification accuracy rate, Kappa coefficient and mean F-measure value of the Morphological method were found as 87.11%, 0.7369, 0.869 for the healthy and 78.08%, 0.5715, 0.782 for the patient, respectively. For the second database, the same statistical measures were determined as 77.02%, 0.4308, 0.755 for the healthy and 76.79%, 0.5227, 0.759 for the patient, respectively. The performance results of the study, which is consistent with real life applications, were compared with the previous studies in this field listed in the literature.

A New Approach to Diagnosing of Importance Degree of Obstructive Sleep Apnea Syndrome: Pairwise AIRS and Fuzzy-AIRS Classifiers

Artificial Immune Recognition System (AIRS) classifier algorithm is robust and effective in medical dataset classification applications such as breast cancer, heart disease, diabetes diagnosis etc. In our previous work, we have proposed a new resource allocation mechanism called fuzzy resource allocation in AIRS algorithm both to improve the classification accuracy and to decrease the computation time in classification process. Here, AIRS and Fuzzy-AIRS classifier algorithms and one against all approach have been combined to increase the classification accuracy of obstructive sleep apnea syndrome (OSAS) that is an important disease that influences both the right and the left cardiac ventricle. The OSAS dataset consists of four classes including of normal (25 subjects), mild OSAS (AHI (Apnea and Hypoapnea Index) =5-15 and 14 subjects), moderate OSAS (AHI < 15-30 and 18 subjects), and serious OSAS (AHI > 30 and 26 subjects). In the extracting of features that is characterized the OSAS disease, the clinical features obtained from Polysomnography used diagnostic tool for obstructive sleep apnea in patients clinically suspected of suffering from this disease have been used. The used clinical features are Arousals Index (ARI), Apnea and Hypoapnea Index (AHI), SaO2 minimum value in stage of REM, and Percent Sleep Time (PST) in stage of SaO2 intervals bigger than 89%. Even though AIRS and Fuzzy-AIRS classifiers have been used in the classifying multi-class problems, theirs classification performances are low in the case of multi-class classification problems. Therefore, we have used two classes in AIRS and Fuzzy-AIRS classifiers by means of one against all approach instead of four classes comprising the healthy subjects, mild OSAS, moderate OSAS, and serious OSAS. We have applied the AIRS, Fuzzy-AIRS, AIRS with one against all approach (Pairwise AIRS), and Fuzzy-AIRS with one against all approach (Pairwise Fuzzy-AIRS) to OSAS dataset. The obtained classification accuracies are 63.41%, 63.41%, 87.19%, and 84.14% using the above methods for 200 resources, respectively. These results show that the best method for diagnosis of OSAS is the combination of AIRS and one against all approach (Pairwise AIRS).

Detection of the electrode disconnection in sleep signals

Sleep staging process that is performed in sleep laboratories in hospitals has an important role in diagnosing some of the sleep disorders and disturbances which are seen in sleep. And also it is an indispensable method. It is usually performed by a sleep expert through examining during the night of the patients (6-8 hours) recorded Electroencephalogram (EEG), Electrooculogram (EOG), Electromyogram (EMG), electrocardiogram (ECG) and other some signals of the patients and determining the stages of sleep in different time sections named as epochs. Manual sleep staging is preferred among the sleep experts but because it is rather tiring and time consuming task, automatic sleep stage scoring studies has come to the fore. However, none of the so far made automatic sleep staging was not accepted by the experts. The most important reason is that the results of the automated systems are not desired accuracy. There are many factors that affecting the accuracy of the systems, such as noise, the inter-channel interference, excessive body movements and disconnection of electrodes. In this study, we examined the written an algorithm to be able to determine to what extent the disconnection of electrodes in EEG signal that obtained one healthy person at the sleep laboratory of Meram Medicine Faculty of Necmettin Erbakan University. According to the obtained application results, the electrodes disconnection in EEG signal could be detected maximum of 100% and minimum of 99.12% accuracy. Accordingly, based on the success achieved in the study, this algorithm is thought to contribute positively to the researchers that the work on and will work on sleep staging problems and increase the success of automatic sleep staging systems.

Comparison of some spectral analysis methods in detection of sleep spindles using YSA

Sleep spindle is a very determinant factor for detection of Non-REM2 stage in sleep staging studies. When it is considered that about half of the sleep consists of Non-REM2 stage, the importance of automatic sleep spindle detection stands out. In this study, three different spectral analysis method- FFT, Welch and AR have been used to estimate the frequency spectrum of sleep EEG signal and feature extraction from this spectrum has been realized. Obtained features have been used in ANN to classify EEG epochs as epochs with spindle and epochs without spindle. It has been observed that least classification error was obtained with FFT as 15.16%.

New Data Pre-processing on Assessing of Obstructive Sleep Apnea Syndrome: Line Based Normalization Method (LBNM)

Sleep disorders are a very common unawareness illness among public. Obstructive Sleep Apnea Syndrome (OSAS) is characterized with decreased oxygen saturation level and repetitive upper respiratory tract obstruction episodes during full night sleep. In the present study, we have proposed a novel data normalization method called Line Based Normalization Method (LBNM) to evaluate OSAS using real data set obtained from Polysomnography device as a diagnostic tool in patients and clinically suspected of suffering OSAS. Here, we have combined the LBNM and classification methods comprising C4.5 decision tree classifier and Artificial Neural Network (ANN) to diagnose the OSAS. Firstly, each clinical feature in OSAS dataset is scaled by LBNM method in the range of [0,1]. Secondly, normalized OSAS dataset is classified using different classifier algorithms including C4.5 decision tree classifier and ANN, respectively. The proposed normalization method was compared with min-max normalization, z-score normalization, and decimal scaling methods existing in literature on the diagnosis of OSAS. LBNM has produced very promising results on the assessing of OSAS. Also, this method could be applied to other biomedical datasets.