Time-frequency analysis of accelerometry data for detection of myoclonic seizures.

Four time-frequency and time-scale methods are studied for their ability of detecting myoclonic seizures from accelerometric data. Methods that are used are: the short-time Fourier transform (STFT), the Wigner distribution (WD), the continuous wavelet transform (CWT) using a Daubechies wavelet, and a newly introduced model-based matched wavelet transform (MOD). Real patient data are analyzed using these four time-frequency and time-scale methods. To obtain quantitative results, all four methods are evaluated in a linear classification setup. Data from 15 patients are used for training and data from 21 patients for testing. Using features based on the CWT and MOD, the success rate of the classifier was 80%. Using STFT or WD-based features, the classification success is reduced. Analysis of the false positives revealed that they were either clonic seizures, the onset of tonic seizures, or sharp peaks in "normal" movements indicating that the patient was making a jerky movement. All these movements are considered clinically important to detect. Thus, the results show that both CWT and MOD are useful for the detection of myoclonic seizures. On top of that, MOD has the advantage that it consists of parameters that are related to seizure duration and intensity that are physiologically meaningful. Furthermore, in future work, the model can also be useful for the detection of other motor seizure types.

An investigation of the phase locking index for measuring of interdependency of cortical source signals recorded in the EEG.

The phase locking index (PLI) was introduced to quantify in a statistical sense the phase synchronization of two signals. It has been commonly used to process biosignals. In this article, we investigate the PLI for measuring the interdependency of cortical source signals (CSSs) recorded in the Electroencephalogram (EEG). To this end, we consider simple analytical models for the mapping of simulated CSSs into the EEG. For these models, the PLI is investigated analytically and through numerical simulations. An evaluation is made of the sensitivity of the PLI to the amount of crosstalk between the sources through biological tissues of the head. It is found that the PLI is a useful interdependency measure for CSSs, especially when the amount of crosstalk is small. Another common interdependency measure is the coherence. A direct comparison of both measures has not been made in the literature so far. We assess the performance of the PLI and coherence for estimation and detection purposes based on, respectively, a normalized variance and a novel statistical measure termed contrast. Based on these performance measures, it is found that the PLI is similar or better than the CM in most cases. This result is also confirmed through analysis of EEGs recorded from epileptic patients.

Detection of subtle nocturnal motor activity from 3-D accelerometry recordings in epilepsy patients.

This paper presents a first step towards reliable detection of nocturnal epileptic seizures based on 3-D accelerometry (ACM) recordings. The main goal is to distinguish between data with and without subtle nocturnal motor activity, thus reducing the amount of data that needs further (more complex) analysis for seizure detection. From 15 ACM signals (measured on five positions on the body), two features are computed, the variance and the jerk. In the resulting 2-D feature space, a linear threshold function is used for classification. For training and testing, the algorithm ACM data along with video data is used from nocturnal registrations in seven mentally retarded patients with severe epilepsy. Per patient, the algorithm detected 100% of the periods of motor activity that are marked in video recordings and the ACM signals by experts. From all the detections, 43%-89% was correct (mean =65%). We were able to reduce the amount of data that need to be analyzed considerably. The results show that our approach can be used for detection of subtle nocturnal motor activity. Furthermore, our results indicate that our algorithm is robust for fluctuations across patients. Consequently, there is no need for training the algorithm for each new patient.

Separation of spatially localized cortical activities in the EEG: model-based analysis and optimization.

The mapping of brain sources into the scalp electroencephalogram (EEG) depends on volume conduction properties of the head and on an electrode montage involving a reference. In this article, the source mapping (SM) is formalized mathematically in the form of an observation function (OF) matrix. The OF-matrix is used to analyze and optimize the SM for a generation model for the desynchronized spontaneous EEG. The optimization leads to a novel reference that minimizes the impact in the EEG of the sources located distant from the electrodes. Thereby, this reference separates spatially localized cortical activities in the EEG. For this reason, it is called the localized reference (LR). The LR is compared with the Hjorth Laplacian reference (HR), which is commonly used for recordings of localized cortical activities. The comparison is made in terms of the relative power contribution of the sources into EEG channels. For the model, the LR is found to have up to 15-20% better performance than the HR, and thus the LR is considered a good alternative to the HR when a head model is available. The HR is, however, a fair approximation of the LR and thus is close to optimum for practical intents and purposes.

Analysis of the phase locking index for measuring of interdependency of cortical signals recorded in the EEG.

The phase locking index (PLI) was introduced to quantify in a statistical sense the phase synchronization of two signals. It has been commonly used to process biosignals. In this paper, we analyze the PLI for measuring the interdependency of cortical source signals (CSSs) recorded in the Electroencephalogram (EEG). The main focus of the analysis is the probability density function, which describes the sensitivity of the PLI to the joint noise ensemble in the CSSs. Since this function is mathematically intractable, we derive approximations and analyze them for a simple analytical model of the CSS mixture in the EEG. The accuracies of the approximate probability density functions (APDFs) are evaluated using simulations for the model. The APDFs are found sufficiently accurate and thus are applicable for practical intents and purposes. They can hence be used to determine the confidence intervals and significance levels for detection methods for interdependencies, e.g., between cortical signals recorded in the EEG.

A model of heart rate changes to detect seizures in severe epilepsy.

AIM OF THE STUDY: An explorative study to assess the value of a model for the automatic detection and characterization of heart rate (HR) changes during seizures in severe epilepsy. METHODS: Heart rate changes were monitored in 10 patients with 104 seizures, mostly tonic and myoclonic, to assess the value of various modalities for the detection of seizures based on heart rate. EEG/video monitoring served as the golden standard. Two algorithms were developed. First, a curve-fitting algorithm was used to characterize the heart rate patterns. A second algorithm based on a moving median filter was developed for automatic detection of the heart rate change onset. For varying model parameters the sensitivity (SENS) and positive predictive values (PPV) were determined. RESULTS: Changes in heart rate were found in 8 of the 10 patients and 50 of 104 seizures. Patterns of heart rate changes could be quantitatively characterized and were found to be stereotype for each individual patient. Large differences of the curve-fitting pattern were in some cases due to a tachycardia at seizure onset that was followed by a significant postictal bradycardia. In two out of three patients with more than 10 seizures a PPV of at least 50% yielded a SENS above 90%. CONCLUSIONS: Heart rate patterns can be accurately characterized with a new developed curve-fitting algorithm. Heart rate changes can also be used for automatic detection of seizures in patients with severe epilepsy if the model parameters are chosen according to predefined characteristics of the patient.

The potential value of three-dimensional accelerometry for detection of motor seizures in severe epilepsy.

Seizure detection results based on the visual analysis of three-dimensional (3D) accelerometry (ACM) and video/EEG recordings are reported for 18 patients with severe epilepsy. They were monitored for 36 hours during which 897 seizures were detected. This was seven times higher than the number of seizures reported by nurses during the registration period. The results in this article demonstrate that 3D ACM is a valuable sensing method for seizure detection in this population. Four hundred twenty-eight (48%) seizures were detected by ACM. With 3D ACM alone it was possible to detect all the seizures in 10 of the 18 patients. Three-dimensional ACM also was complementary to EEG in our population. ACM patterns during seizures were stereotypical in 95% of the motor seizures. These characteristic patterns are a starting point for automated seizure detection.