Comparison of fractal dimension estimation algorithms for epileptic seizure onset detection.

Fractal dimension (FD) is a natural measure of the irregularity of a curve. In this study the performances of three waveform FD estimation algorithms (i.e. Katz's, Higuchi's and the k-nearest neighbour (k-NN) algorithm) were compared in terms of their ability to detect the onset of epileptic seizures in scalp electroencephalogram (EEG). The selection of parameters involved in FD estimation, evaluation of the accuracy of the different algorithms and assessment of their robustness in the presence of noise were performed based on synthetic signals of known FD. When applied to scalp EEG data, Katz's and Higuchi's algorithms were found to be incapable of producing consistent changes of a single type (either a drop or an increase) during seizures. On the other hand, the k-NN algorithm produced a drop, starting close to the seizure onset, in most seizures of all patients. The k-NN algorithm outperformed both Katz's and Higuchi's algorithms in terms of robustness in the presence of noise and seizure onset detection ability. The seizure detection methodology, based on the k-NN algorithm, yielded in the training data set a sensitivity of 100% with 10.10 s mean detection delay and a false positive rate of 0.27 h(-1), while the corresponding values in the testing data set were 100%, 8.82 s and 0.42 h(-1), respectively. The above detection results compare favourably to those of other seizure onset detection methodologies applied to scalp EEG in the literature. The methodology described, based on the k-NN algorithm, appears to be promising for the detection of the onset of epileptic seizures based on scalp EEG.

Time-frequency analysis methods to quantify the time-varying microstructure of sleep EEG spindles: possibility for dementia biomarkers?

The time-varying microstructure of sleep EEG spindles may have clinical significance in dementia studies and can be quantified with a number of techniques. In this paper, real and simulated sleep spindles were regarded as AM/FM signals modeled by six parameters that define the instantaneous envelope (IE) and instantaneous frequency (IF) waveforms for a sleep spindle. These parameters were estimated using four different methods, namely the Hilbert transform (HT), complex demodulation (CD), matching pursuit (MP) and wavelet transform (WT). The average error in estimating these parameters was lowest for HT, higher but still less than 10% for CD and MP, and highest (greater than 10%) for WT. The signal distortion induced by the use of a given method was greatest in the case of HT and MP. These two techniques would necessitate the removal of about 0.4s from the spindle data, which is an important limitation for the case of spindles with duration less than 1s. Although the CD method may lead to a higher error than HT and MP, it requires a removal of only about 0.23s of data. An application of this sleep spindle parameterization via the CD method is proposed, in search of efficient EEG-based biomarkers in dementia. Preliminary results indicate that the proposed parameterization may be promising, since it can quantify specific differences in IE and IF characteristics between sleep spindles from dementia subjects and those from aged controls.

Potential dementia biomarkers based on the time-varying microstructure of sleep EEG spindles.

The time-varying microstructure of sleep EEG spindles may have clinical significance in dementia studies. In this work, the sleep spindle is modeled as an AM-FM signal and parameterized in terms of six parameters, three quantifying the instantaneous envelope (IE) and three quantifying the instantaneous frequency (IF) of the spindle model. The IE and IF waveforms of sleep spindles from patients with dementia and normal controls were estimated using the time-frequency technique of Complex Demodulation (CD). Sinusoidal curve-fitting using a matching pursuit (MP) approach was applied to the IE and IF waveforms for the estimation of the six model parameters. Specific differences were found in sleep spindle instantaneous frequency dynamics between spindles from dementia subjects and spindles from controls.

Modeling the time-varying microstructure of simulated sleep EEG spindles using time-frequency analysis methods.

The time-varying microstructure of sleep spindles may have clinical significance and can be quantified and modeled with a number of techniques. In this paper, sleep spindles were regarded as AM-FM signals modeled by six parameters. The instantaneous envelope (IE) and instantaneous frequency (IF) waveforms were estimated using four different methods, namely Hilbert Transform (HT), Complex Demodulation (CD), Wavelet Transform (WT) and Matching Pursuit (MP). The six model parameters were subsequently estimated from the IE and IF waveforms. The average error, taking into account the error for each model parameter, was lowest for HT, higher but still less than 10% for CD and MP, and highest (greater than 10%) for WT, for three different spindle model examples. The amount of distortion induced by the use of a given method is also important; distortion was the greatest (0.4 sec) in the case of HT. Therefore, in the case of real spindles, one could utilize CD and MP and, if the spindle duration is more than 1 sec, HT as well.

Quantifying and visualizing uncertainty in EEG data of neonatal seizures.

This paper presents an approach to quantifying and visualizing uncertainty in EEG data of neonatal seizures. This approach exploits the inherent ability of trained quantum neural networks (QNNs) to learn arbitrary membership profiles from sample data. The ability of QNNs to quantify uncertainty in data is combined with the ability of ordered self-organizing maps (SOMs) to recognize structure in data and allow its visualization in two dimensions. The proposed approach is evaluated using EEG data of neonates monitored for seizures.

Sleep spindle incidence dynamics: a pilot study based on a Markovian analysis.

Results are reported, based on 5 healthy subjects, concerning patterns in the dynamics of the sequential arrangement of spindles in human stage 2 sleep. Specifically, the conditional probability of incidence of successive spindle lengths and successive inter-spindle intervals is investigated. The results show that successive spindle lengths may be statistically independent. However, their distribution (histogram) may be similar for two different stage 2 periods, one in the first third and another in the second third of the night sleep record. In contrast to the finding about spindle lengths, results show that successive inter-spindle intervals may not be statistically independent. Furthermore, the overall dynamics of the sequential arrangement of inter-spindle intervals may be similar for the two sleep periods. These findings are discussed in the context of the "sleep maintenance" role of spindles.

An automated system for epileptogenic focus localization in the electroencephalogram.

This paper describes an automated system for the detection and localization of foci of epileptiform activity in the EEG. The system detects sharp EEG transients (STs) in the process, but the emphasis is on epileptic focus localization. A combination of techniques involving signal processing, pattern recognition, and the expert rules of an experienced electroencephalographer, involving considerable spatiotemporal context information, is applied to multichannel EEG data. An overall emphasis on minimizing the number of false-positive sharp transient detections drives the system design. Tested on data from 13 subjects with epileptiform activity and 5 controls, all areas of focal epileptiform activity were detected by the system, although not all of the contributing foci were reported separately. Two false-positive foci were detected as well due to nonfocal spike activity and normal spike-like activity not present in the training set. The system detected 95.7% of the epileptiform events constituting the correctly detected foci, with a false detection rate of 11.1%.

Computer-based recognition of EEG patterns.

A critical "mini-overview" is presented of several computer-based techniques proposed for the automated recognition of specific EEG patterns, important in visual EEG analysis. Both phasic and tonic EEG patterns are addressed. The techniques discussed include methods based on power spectrum analysis and on period-amplitude analysis, "mimetic" methods and related implementations in an expert system approach, and methods based on artificial neural networks.

Delta (0.5-1.5 Hz) and sigma (11.5-15.5 Hz) EEG power dynamics throughout quiet sleep in infants.

Twenty-eight healthy infants, split into 3 groups according to age (group 1: 2-6 weeks, n = 10; group 2: 7-14 weeks, n = 10; and group 3: 4-12 months, n = 8), were recorded during the whole night. For each infant, the longest quiet sleep (QS) phase occurring between 8 p.m. and midnight was selected for EEG power spectral analysis. The power in the frequency band related to low-frequency delta waves (0.5-1.5 Hz, "delta band") and the power in the frequency band related to sigma spindles (11.5-15.5 Hz, "sigma band") were analyzed. Group 1 infants showed no significant modification of the power in the sigma band in the course of the QS phase; the power in the delta band showed a significant increase between the second and the third 5 min segment and a decrease thereafter. Group 2 infants showed a progressive reduction of the power in the sigma band, whereas the power in the delta band increased during the first 15 min. In group 3 infants, the sigma band power significantly decreased between the third and the fifth 5 min segment without further changes. The power in the delta band, on the contrary, increased progressively for the first 20 min and showed a second progressive increase beyond 35 min. For both group 2 and group 3 infants, it appeared that the change in delta power preceded the change in sigma power. The above results provide quantitative evidence that a well-defined temporal inhomogeneity pattern in the EEG of the QS phase may appear between 7 and 14 weeks of age and continues from the fourth month on.

Estimation of time delay between EEG signals for epileptic focus localization: statistical error considerations.

A theoretical analysis of the variance for the time delay estimate between two EEG signals, obtained via the phase spectrum method, is presented. Explicit theoretical formulae for the variance are obtained and compared via simulations to experimentally derived results for estimate variability. The variance of the time delay estimate is inversely proportional to the frequency range of interest, to the number of data segments utilized for spectral estimation, and to the coherence between the two EEG signals. The simulations indicate that the formulae can be used even with non-gaussian and relatively narrow-band EEG-like data. A minimum-variance estimate for the time delay is also presented.

Developmental changes in the clustering pattern of sleep rapid eye movement activity during the first year of life: a Markov-process approach.

Findings are presented in support of the hypothesis that the tendency of sleep rapid eye movement (REM) activity to group into burst structures changes with age during the first year of life in normal infants. Specifically, by assuming a markovian model for the generation of 1 sec long units of REM activity, it is shown that the propensity of those units to develop a sustained clustering pattern may increase during the first 2 months, possibly reaching a plateau at about 4 months. On the other hand, the overall density of REM activity units may continue to increase beyond that point in time.

Knowledge-based interpretation of bioelectrical signals.

The integration into one automated system of the qualitative knowledge of the physician, who is accustomed to visually scanning a paper record, with possibly sophisticated signal analysis tools that must replace the visual scanning is addressed. The automated interpretation of bioelectrical signals must combine signal analysis knowledge with knowledge about the application domain of the signal and, as such, requires the knowledge of two experts. The role of event-oriented interpretation and the use of spatial and temporal context is discussed. An architecture for a system tailored to fit the characteristics of automated bioelectrical signal interpretation is described. It contains two separate, cooperating knowledge bases, and it makes use of object-oriented knowledge representation and spatial and temporal operators. An example of the automated detection of epileptogenic electroencephalograph sharp transients is presented.

Context-based automated detection of epileptogenic sharp transients in the EEG: elimination of false positives.

This paper describes a knowledge-based system for the elimination of false positives in the automated detection of epileptogenic sharp transients in the EEG. The system makes comprehensive use of spatial and temporal context information available on 16 channels of EEG, EKG, EMG, and EOG. A knowledge-based implementation is used because of the ease with which it allows the contextual rules to be expressed and refined. The resulting system is shown to be capable of rejecting a wide variety of artifacts commonly found in EEG recordings, artifacts that cause numerous false positive detections in systems making less comprehensive use of context.

Non-random patterns of REM occurrences during REM sleep in normal human subjects: an automated second-order study using Markovian modeling.

An automated analysis of the patterns in REM occurrences during REM sleep in 6 healthy young adults was performed, with an emphasis on second-order parameters. It was found that the majority of REMs were grouped in bursts with a tendency to return to the burst mode once outside of it. The occurrence pattern of REMs within bursts was found not to be of a purely random (renewal) nature, in distinction to that of isolated REMs. Second-order REM occurrence patterns, quantified via Markovian modeling, were not stationary when REM period segments of less than 8 min duration were considered, and those patterns remained fairly constant from REM period to REM period within the night. First-order parameters and non-Markovian second-order parameters showed a less stable behaviour throughout the night. It is concluded that there may exist 2 aspects to REM generation, a relatively unstable (i.e., variable) phasic aspect, quantified by first-order parameters, and a more stable tonic aspect, quantified by second- and possibly higher-order parameters.

Automated analysis of abnormal electroencephalograms.

This paper presents a critical review of various attempts at computerized analysis of abnormal electroencephalograms (EEGs). A description of normal and abnormal EEGs from the viewpoint of the clinician is presented at first, along with guidelines used in the visual detection and quantification of EEG abnormalities, followed by a brief review of some important computerized methodologies for clinical EEG analysis. Automated detection and quantification of epileptogenic EEG transients and seizures are reviewed next, and digital computer (software) as well as hardwired systems are presented. Computerized techniques for the quantification of abnormal EEGs in cerebrovascular disorders and coma, metabolic disorders, and for the localization of brain lesions and tumors are presented as well. Future directions and the general problem of man-machine agreement are elaborated upon.

Computer-aided quantification of EEG spike and sharp wave characteristics.

This work presents data from detailed, computer-aided analysis of pertinent electrographic characteristics of well-defined EEG spikes and sharp waves. The data show morphological differences between spikes obtained from different subjects, spikes from different electrode montages, as well as between monophasic and biphasic spikes, and between spikes and sharp waves.