Automatic detection of EEG artefacts arising from head movements using EEG and gyroscope signals.

Contamination of EEG signals by artefacts arising from head movements has been a serious obstacle in the deployment of automatic neurological event detection systems in ambulatory EEG. In this paper, we present work on categorizing these head-movement artefacts as one distinct class and on using support vector machines to automatically detect their presence. The use of additional physical signals in detecting head-movement artefacts is also investigated by means of support vector machines classifiers implemented with gyroscope waveforms. Finally, the combination of features extracted from EEG and gyroscope signals is explored in order to design an algorithm which incorporates both physical and physiological signals in accurately detecting artefacts arising from head-movements.

Dynamic, location-based channel selection for power consumption reduction in EEG analysis.

The objective of this study is to develop methods to dynamically select EEG channels to reduce power consumption in seizure detection while maintaining detection accuracy. A method is proposed whereby a number of primary screening channels are predefined. Depending on the classification results of those channels, further channels are selected for analysis. This method provides savings in computational complexity of 43%. A further method called idling is then proposed which increases the computational saving to 75%. The performance of a location-independent, decision-based method is used for comparison. The proposed method achieves better computational savings for the same performance than the decision-based method. The decision-based method was capable of higher overall computational savings, but with a reduction in seizure detection performance. Each method was also implemented with the REACT algorithm on a Blackfin microprocessor and the average power measured. The proposed methods gave a power saving of up to 47% with no reduction in detection performance.

Automatic detection of EEG artefacts arising from head movements.

The need for reliable detection of artefacts in raw and processed EEG is widely acknowledged. In this paper, we present the results of an investigation into appropriate features for artefact detection in the REACT ambulatory EEG system. The study focuses on EEG artefacts arising from head movement. The use of one generalised movement artefact class to detect movement artefacts is proposed. Temporal, frequency, and entropy-based features are evaluated using Kolmogorov-Smirnov and Wilcoxon rank-sum non-parametric tests, Mutual Information Evaluation Function and Linear Discriminant Analysis. Results indicate good separation between normal EEG and artefacts arising from head movement, providing a strong argument for treating these head movement artefacts as one generalised class rather than treating their component signals individually.

Dynamic channel selection to reduce computational burden in seizure detection.

Ambulatory physiological monitoring devices benefit patients, medical staff and hospitals by allowing patients to return home with the devices for monitoring. The main problem associated with designing such devices is that of power consumption. Wireless communications and complex processing are generally part of such devices and are power hungry components. These problems are magnified when dealing with EEG signals, with relatively high data rates, multiple channels, and advanced signal processing techniques required. This paper proposes a method to dynamically select EEG channels in the REACT seizure detection system based on information already available in the system, hence keeping any added computational complexity very low. Using the techniques computational effort can be reduced by up to 65% with no effect on the REACT seizure detection performance.

EEG compression using JPEG2000: how much loss is too much?

Compression of biosignals is an important means of conserving power in wireless body area networks and ambulatory monitoring systems. In contrast to lossless compression techniques, lossy compression algorithms can achieve higher compression ratios and hence, higher power savings, at the expense of some degradation of the reconstructed signal. In this paper, a variant of the lossy JPEG2000 algorithm is applied to Electroencephalogram (EEG) data from the Freiburg epilepsy database. By varying compression parameters, a range of reconstructions of varying signal fidelity is produced. Although lossy compression has been applied to EEG data in previous studies, it is unclear what level of signal degradation, if any, would be acceptable to a clinician before diagnostically significant information is lost. In this paper, the reconstructed EEG signals are applied to REACT, a state-of-the-art seizure detection algorithm, in order to determine the effect of lossy compression on its seizure detection ability. By using REACT in place of a clinician, many hundreds of hours of reconstructed EEG data are efficiently analysed, thereby allowing an analysis of the amount of EEG signal distortion that can be tolerated. The corresponding compression ratios that can be achieved are also presented.

Ambulatory REACT: real-time seizure detection with a DSP microprocessor.

REACT (Real-Time EEG Analysis for event deteCTion) is a Support Vector Machine based technology which, in recent years, has been successfully applied to the problem of automated seizure detection in both adults and neonates. This paper describes the implementation of REACT on a commercial DSP microprocessor; the Analog Devices Blackfin®. The primary aim of this work is to develop a prototype system for use in ambulatory or in-ward automated EEG analysis. Furthermore, the complexity of the various stages of the REACT algorithm on the Blackfin processor is analysed; in particular the EEG feature extraction stages. This hardware profile is used to select a reduced, platform-aware feature set, in order to evaluate the seizure classification accuracy of a lower-complexity, lower-power REACT system.

Age-independent seizure detection.

This paper examines whether an appropriate algorithm, developed for use with neonatal data, could also be used, without alteration, for the detection of seizures in adults with epilepsy. The performance of a feature extraction and SVM classifier system is evaluated on databases of 17 neonatal patients and 15 adult patients. Mean ROC curve areas of 0.96 and 0.94 for neonatal and adult databases respectively show that high accuracy can be achieved independent of age. It is also shown that features contribute differently for neonatal and adult data.

Energy-efficient low duty cycle MAC protocol for wireless body area networks.

This paper presents an energy-efficient medium access control protocol suitable for communication in a wireless body area network for remote monitoring of physiological signals such as EEG and ECG. The protocol takes advantage of the static nature of the body area network to implement the effective time-division multiple access (TDMA) strategy with very little amount of overhead and almost no idle listening (by static, we refer to the fixed topology of the network investigated). The main goal is to develop energy-efficient and reliable communication protocol to support streaming of large amount of data. TDMA synchronization problems are discussed and solutions are presented. Equations for duty cycle calculation are also derived for power consumption and battery life predictions. The power consumption model was also validated through measurements. Our results show that the protocol is energy efficient for streaming communication as well as sending short bursts of data, and thus can be used for different types of physiological signals with different sample rates. The protocol is implemented on the analog devices ADF7020 RF transceivers.

Gaussian process modeling of EEG for the detection of neonatal seizures.

Gaussian process (GP) probabilistic models have attractive advantages over parametric and neural network modeling approaches. They have a small number of tuneable parameters, can be trained on relatively small training sets, and provide a measure of prediction certainty. In this paper, these properties are exploited to develop two methods of highlighting the presence of neonatal seizures from electroencephalograph (EEG) signals. In the first method, the certainty of the GP model prediction is used to indicate the presence of seizures. In the second approach, the hyperparameters of the GP model are used. Tests are carried out with a feature set of ten EEG measures developed from various signal processing techniques. Features are evaluated using a neural network classifier on 51 h of real neonatal EEG. The GP measures, in particular, the prediction certainty approach, produce a high level of performance compared to other modeling methods and methods currently in clinical use for EEG analysis, indicating that they are an important and useful tool for the real-time detection of neonatal seizures.

An evaluation of automated neonatal seizure detection methods.

OBJECTIVE: To evaluate 3 published automated algorithms for detecting seizures in neonatal EEG. METHODS: One-minute, artifact-free EEG segments consisting of either EEG seizure activity or non-seizure EEG activity were extracted from EEG recordings of 13 neonates. Three published neonatal seizure detection algorithms were tested on each EEG recording. In an attempt to obtain improved detection rates, threshold values in each algorithm were manipulated and the actual algorithms were altered. RESULTS: We tested 43 data files containing seizure activity and 34 data files free from seizure activity. The best results for Gotman, Liu and Celka, respectively, were as follows: sensitivities of 62.5, 42.9 and 66.1% along with specificities of 64.0, 90.2 and 56.0%. CONCLUSIONS: The levels of performance achieved by the seizure detection algorithms are not high enough for use in a clinical environment. The algorithm performance figures for our data set are considerably worse than those quoted in the original algorithm source papers. The overlap of frequency characteristics of seizure and non-seizure EEG, artifacts and natural variances in the neonatal EEG cause a great problem to the seizure detection algorithms. SIGNIFICANCE: This study shows the difficulties involved in detecting seizures in neonates and the lack of a reliable detection scheme for clinical use. It is clear from this study that while each algorithm does produce some meaningful information, the information would only be usable in a reliable neonatal seizure detection process when accompanied by more complex analysis, and more advanced classifiers.