Nils Löfgren

Automatic classification of background EEG activity in healthy and sick neonates.

The overall aim of our research is to develop methods for a monitoring system to be used at neonatal intensive care units. When monitoring a baby, a range of different types of background activity needs to be considered. In this work, we have developed a scheme for automatic classification of background EEG activity in newborn babies. EEG from six full-term babies who were displaying a burst suppression pattern while suffering from the after-effects of asphyxia during birth was included along with EEG from 20 full-term healthy newborn babies. The signals from the healthy babies were divided into four behavioural states: active awake, quiet awake, active sleep and quiet sleep. By using a number of features extracted from the EEG together with Fishers linear discriminant classifier we have managed to achieve 100% correct classification when separating burst suppression EEG from all four healthy EEG types and 93% true positive classification when separating quiet sleep from the other types. The other three sleep stages could not be classified. When the pathological burst suppression pattern was detected, the analysis was taken one step further and the signal was segmented into burst and suppression, allowing clinically relevant parameters such as suppression length and burst suppression ratio to be calculated. The segmentation of the burst suppression EEG works well, with a probability of error around 4%.

Classification of burst and suppression in the neonatal electroencephalogram

Fishers linear discriminant (FLD), a feed-forward artificial neural network (ANN) and a support vector machine (SVM) were compared with respect to their ability to distinguish bursts from suppressions in electroencephalograms (EEG) displaying a burst-suppression pattern. Five features extracted from the EEG were used as inputs. The study was based on EEG signals from six full-term infants who had suffered from perinatal asphyxia, and the methods have been trained with reference data classified by an experienced electroencephalographer. The results are summarized as the area under the curve (AUC), derived from receiver operating characteristic (ROC) curves for the three methods. Based on this, the SVM performs slightly better than the others. Testing the three methods with combinations of increasing numbers of the five features shows that the SVM handles the increasing amount of information better than the other methods.

Spectral analysis of burst periods in EEG from healthy and post-asphyctic full-term neonates

OBJECTIVE To investigate whether the periodic EEG patterns seen in healthy and sick full term neonates (trace alternant and burst suppression, respectively) have different frequency characteristics. METHODS Burst episodes were selected from the EEGs of 9 healthy and 9 post-asphyctic full-term neonates and subjected to power spectrum analysis. Powers in two bands were estimated; 0-4 and 4-30 Hz, designated low- and high-frequency activity, respectively (LFA, HFA). The spectral edge frequency (SEF) was also assessed. RESULTS In bursts, the LFA power was lower in periods of burst suppression as compared to those of trace alternant. The parameter that best discriminated between the groups was the relative amount of low- and high-frequency activity. The SEF parameter had a low sensitivity to the group differences. In healthy neonates, the LFA power was higher over the posterior right as compared to the posterior left region. CONCLUSIONS Spectral power of low frequencies differs significantly between the burst episodes of healthy and sick neonates. SIGNIFICANCE These results can be used when monitoring cerebral function in neonates.

Infraslow EEG activity in burst periods from post asphyctic full term neonates.

OBJECTIVE To investigate whether very low EEG frequency activity can be recorded from post asphyctic full term neonates using EEG equipment where the high pass filter level was lowered to 0.05 Hz. METHODS The time constant of the amplifier hardware was set to 3.2 s in order to enable recordings that equal to a high pass filter cut off at 0.05 Hz. Burst episodes were selected from the EEGs of 5 post asphyctic full term neonates. The episodes were analysed visually using different montages and subjected to power spectrum analysis. Powers in two bands were estimated; 0-1 and 1-4 Hz, designated very low- and low-frequency activity, respectively (VLFA, LFA). RESULTS In all infants, VLFA coinciding with the burst episodes could be detected. The duration of the VLFA was about the same as that of the burst episode i.e. around 4s. The activity was most prominent over the posterior regions. In this small material, a large amount of VLFA neonatally seemed to possibly be related to a more favourable prognosis. CONCLUSIONS VLFA can be recorded from post asphyctic full term neonates using EEG equipment with lowered cut off frequency for the high pass filter. SIGNIFICANCE VLFA normally disregarded due to filtering, is present in the EEG of sick neonates and may carry important clinical information.

A Comparative Study of Fetal Heart Rate Variability Analysis Techniques

This study examines a novel methodology for continuous fetal heart rate variability (FHRV) assessment in a non-stationary intrapartum fetal heart rate (FHR). The specific aim was to investigate simple statistics, dimension estimates and entropy estimates as methods to discriminate situations of low FHRV related to non-reassuring fetal status or as a consequence of sedatives given to the mother. Using a t-test it is found that the dimension of the zero set and sample entropy reveal a difference in mean distribution of significance >99%. Thus it may prove possible to build a discriminating system based on either one or a combination of these techniques

Comparison of Three Methods for Classifying Burst and Suppression in the EEG of Post Asphyctic Newborns

Fishers linear discriminant, a feed-forward neural network (NN) and a support vector machine (SVM) are compared with respect to their ability to distinguish bursts from suppression in burst-suppression electroencephalogram (EEG) signals using five features inherent in the EEG as input. The study is based on EEG signals from six full term infants who have suffered from perinatal asphyxia, and the methods have been trained with reference data classified by an experienced electroencephalographer. The results are summarized as area under the curve (AUC) values derived from receiver operating characteristic (ROC) curves for the three methods, and show that the SVM is slightly better than the others, at the cost of a higher computational complexity.

Spectral distance for ARMA models applied to electroencephalogram for early detection of hypoxia.

A novel measure of spectral distance is presented, which is inspired by the prediction residual parameter presented by Itakura in 1975, but derived from frequency domain data and extended to include autoregressive moving average (ARMA) models. This new algorithm is applied to electroencephalogram (EEG) data from newborn piglets exposed to hypoxia for the purpose of early detection of hypoxia. The performance is evaluated using parameters relevant for potential clinical use, and is found to outperform the Itakura distance, which has proved to be useful for this application. Additionally, we compare the performance with various algorithms previously used for the detection of hypoxia from EEG. Our results based on EEG from newborn piglets show that some detector statistics divert significantly from a reference period less than 2 min after the start of general hypoxia. Among these successful detectors, the proposed spectral distance is the only spectral-based parameter. It therefore appears that spectral changes due to hypoxia are best described by use of an ARMA- model-based spectral estimate, but the drawback of the presented method is high computational effort.

Detection of bursts in the EEG of post asphyctic newborns.

Eight features inherent in the electroencephalogram (EEG) have been extracted and evaluated with respect to their ability to distinguish bursts from suppression in burst-suppression EEG. The study is based on EEG from six full term infants who had suffered from lack of oxygen during birth. The features were used as input in a neural network, which was trained on reference data segmented by an experienced electroencephalographer. The performance was then evaluated on validation data for each feature separately and in combinations. The results show that there are significant variations in the type of activity found in burst-suppression EEG from different subjects, and that while one or a few features seem to be sufficient for most patients in this group, some cases require specific combinations of features for good detection to be possible

Remote sessions and frequency analysis for improved insight into cerebral function during pediatric and neonatal intensive care

A project involving recording and analysing EEG together with cardiovascular signals and temperature has been initiated. The aim of this project is to establish difficulties and possibilities involved with implementing a system for remote sessions and analysing EEG in correlation with other physiological signals. One objective is to find indicators of cerebral function during postasphyxia neonatal intensive care and pediatric cardiopulmonary bypass surgery with hypothermia. Remote sessions for joint interpretation have been carried out between pediatricians and clinical neurophysiologists, and EEG has been analyzed using frequency analyzing tools. One result is the discovery of reversible spectral changes coinciding with blood pressure falls, which may indicate loss of autoregulation function. This finding is one outcome from initial use of a system, developed during the project to facilitate communication about, and analysis of the recorded signals. Thus, already from a limited number of remote sessions and the use of basic signal processing techniques, important results have been achieved and better insight has been gained of how cerebral function is affected by cardiopulmonary bypass surgery. In this paper, we present our experiences from introducing a system for remote consultations, and evaluate the use for such a system in the current applications.

Classifying Burst and Suppression in the EEG of Post Asphyctic Newborns using a Support Vector Machine

A support vector machine (SVM) was trained to distinguish bursts from suppression in burst-suppression EEG, using five features inherent in the electro-encephalogram (EEG) as input. The study was based on data from six full term infants who had suffered from perinatal asphyxia, and the machine was trained with reference classifications made by an experienced electroencephalographer. The results show that the method may be useful, but that differences between patients in the data set makes optimization of the system difficult