V. Gerla

Multivariate Analysis of Full-Term Neonatal Polysomnographic Data

Polysomnography (PSG) is one of the most important noninvasive methods for studying maturation of the child brain. Sleep in infants is significantly different from sleep in adults. This paper addresses the problem of computer analysis of neonatal polygraphic signals. We applied methods designed for differentiating three important neonatal behavioral states: quiet sleep, active sleep, and wakefulness. The proportion of these states is a significant indicator of the maturity of the newborn brain in clinical practice. In this study, we used data provided by the Institute for Care of Mother and Child, Prague (12 newborn infants of similar postconceptional age). The data were scored by an experienced physician to four states (wake, quiet sleep, active sleep, movement artifact). For accurate classification, it was necessary to determine the most informative features. We used a method based on power spectral density (PSD) applied to each EEG channel. We also used features derived from electrooculogram (EOG), electromyogram (EMG), ECG, and respiration [pneumogram (PNG)] signals. The most informative feature was the measure of regularity of respiration from the PNG signal. We designed an algorithm for interpreting these characteristics. This algorithm was based on Markov models. The results of automatic detection of sleep states were compared to the ldquosleep profilesrdquo determined visually. We evaluated both the success rate and the true positive rate of the classification, and statistically significant agreement of the two scorings was found. Two variants, for learning and for testing, were applied, namely learning from the data of all 12 newborns and tenfold cross-validation, and learning from the data of 11 newborns and testing on the data from the 12th newborn. We utilized information obtained from several biological signals (EEG, ECG, PNG, EMG, EOG) for our final classification. We reached the final success rate of 82.5%. The true positive rate was 81.8% and the false positive rate was 6.1%. The most important step in the whole process is feature extraction and feature selection. In this process, we used visualization as an additional tool that helped us to decide which features to select. Proper selection of features may significantly influence the success rate of the classification. We made a visual comparison of the computed features with the manual scoring provided by the expert. A hidden Markov model was used for classification. The advantage of this model is that it determines the future behavior of the process by its present state. In this way, it preserves information about temporal development.

Classification of the emotional states based on the EEG signal processing

The paper proposes a method for the classification of EEG signal based on machine learning methods. We analyzed the data from an EEG experiment consisting of affective picture stimuli presentation, and tested automatic recognition of the individual emotional states from the EEG signal using Bayes classifier. The mean accuracy was about 75 percent, but we were not able to select universal features for classification of all subjects, because of inter-individual differences in the signal. We also identified correlation between the classification error and the extroversion-introversion personality trait measured by EPQ-R test. Introverts have lower excitation threshold so we are able to detect the differences in their EEG activity with better accuracy. Furthermore, the use of Kohonens self-organizing map for visualization is suggested and demonstrated on one subject.

PSGLab Matlab toolbox for polysomnographic data processing: Development and practical application

In this paper, we present a Matlab toolbox for processing of multichannel biomedical data, especially polysomnographic signals. It implements signal preprocessing, feature extraction, supervised and unsupervised classification methods and high-level data visualization techniques. These methods allow finding important information hidden in biomedical signals and its suitable interpretation. We investigated the possibility of applying the combination of segmentation, various data representation methods, clustering and classification to the field of sleep, comatose and neonatal EEG. All datasets were provided by cooperating medical partners. The most important step in the whole process is feature extraction and feature selection. In this process we used visualization as an additional tool that helped us to decide which features to select. Proper selection of features may significantly influence the success rate of the classification.

Feature extraction and classification of EEG sleep recordings in newborns

Visual evaluation of long-term EEG recordings is very difficult, time consuming and subjective process. This paper aims to present the research and development of a comprehensive scheme for computer-assisted recognition of behavioral states of sleep in newborns. In clinical practice, the ratio of behavioral states (wakefulness, quiet and active sleep) is used as an important indicator of the brain maturation. Analysis was performed offline, on real clinical data, with the assumption that each EEG channel in recording was independent from others and equally important for analysis and classification. The proposed solution comprises several consecutive steps of signal preprocessing and processing, with focus on segmentation, feature extraction and selection, and classification. Performed classification was based on linear support vector machines and performance was evaluated through cross validation. Obtained results can be used as a reference for developing or enhancing neonatal sleep EEG/PSG classification algorithms.

System Approach to Complex Signal Processing Task

This paper describes methods of automatic analysis and classification of biological signals. Polysomnographic (PSG) recordings encompass a set of heterogeneous biological signals (e.g. EEG, EOG, EMG, ECG, PNG) recorded simultaneously. These signals, especially EEG, are very complex and exhibit nonstationarity and stochasticity. Thus their processing represents a challenging multilevel procedure composed of several methods. Used methods are illustrated on examples of PSG recordings of newborns and sleep recordings of adults and can be applied to similar tasks in other problem domains. Analysis was performed using real clinical data.

Neonatal EEG sleep stages modelling by temporal profiles

The paper deals with the application of the EEG temporal profiles for the neonatal sleep stages modelling. The temporal profiles created by adaptive segmentation and cluster analysis reflect the time structure of the EEG during different periods of sleep. They can be used for neonatal EEG quantification and for the detection of sleep stage changes.

Modeling the Microstructure of Neonatal EEG Sleep Stages by Temporal Profiles

The electroencephalogram (EEG) and neonatal sleep analysis provide sensitive markers of brain maturation and developmental changes in term and preterm newborns. A new method for automatic sleep stages indication was developed. The procedure is based on processing of time profiles using adaptive segmentation and subsequent classification of extracted signal graphoelements. The time profiles, functions of the class membership in the course of time, reflect the dynamic EEG structure and may be used for indication of changes in the neonatal sleep. The method is able to detect the sleep stage changes even in preterm infants´ EEG.

Improvements in Processing of Neonatal Sleep Electroencephalographic Recordings

The ratio of three newborn’s behavioral states – wakefulness, active sleep and quiet sleep – is an important indicator of the maturity of the newborn brain in clinical practice.

Wards Clustering Method for Distinction between Neonatal Sleep Stages

A robust, automated classification system for polysomnographic (PSG) data targeted to the newborn sleep stage identification is presented. The problem of polysomnographic signal classification is very often difficult because of artifacts and noise. Furthermore, for each signal, a special classification method for each particular type of segment must be mostly used. This paper proposes fully unsupervised approach using adaptive segmentation, appropriate features extraction and hierarchical clustering (Ward’s minimumvariance method is used). The mutual information concept was applied to results of hierarchical clustering. The proposed procedure was tested on real neonatal data. All sleep states were successfully separated by a combination of EEG, EMG, EOG, PNG and ECG channels.

Visualization methods used for evaluation of neonatal polysomnographic data

Polysomnographic (PSG) signal processing represents a complex process consisting of several subsequent steps, namely pre-processing, segmentation, extraction of descriptive features, and classification. In this paper we focus on visualization methods that are also unseparable part of the whole process. The aim of these methods is to ease the work of medical doctors and to show trends that are not obvious when performing a manual inspection of the recorded signal. In this study, the designed methods are applied to neonatal PSG data and enable the enhancement in visual differentiation between three important behavioral states: quiet sleep (QS), active sleep (AS) and wakefulness (WK). The ratio of these states is a significant indicator of the maturity of the newborn brain in clinical practice.