Rosana Esteller

Epileptic Seizures May Begin Hours in Advance of Clinical Onset: A Report of Five Patients*

Mechanisms underlying seizure generation are traditionally thought to act over seconds to minutes before clinical seizure onset. We analyzed continuous 3- to 14-day intracranial EEG recordings from five patients with mesial temporal lobe epilepsy obtained during evaluation for epilepsy surgery. We found localized quantitative EEG changes identifying prolonged bursts of complex epileptiform discharges that became more prevalent 7 hr before seizures and highly localized subclinical seizure-like activity that became more frequent 2 hr prior to seizure onset. Accumulated energy increased in the 50 min before seizure onset, compared to baseline. These observations, from a small number of patients, suggest that epileptic seizures may begin as a cascade of electrophysiological events that evolve over hours and that quantitative measures of preseizure electrical activity could possibly be used to predict seizures far in advance of clinical onset.

A comparison of waveform fractal dimension algorithms

The fractal dimension of a waveform represents a powerful tool for transient detection. In particular, in analysis of electroencephalograms and electrocardiograms, this feature has been used to identify and distinguish specific states of physiologic function. A variety of algorithms are available for the computation of fractal dimension. In this study, the most common methods of estimating the fractal dimension of biomedical signals directly in the time domain (considering the time series as a geometric object) are analyzed and compared. The analysis is performed over both synthetic data and intracranial electroencephalogram data recorded during presurgical evaluation of individuals with epileptic seizures. The advantages and drawbacks of each technique are highlighted. The effects of window size, number of overlapping points, and signal-to-noise ratio are evaluated for each method. This study demonstrates that a careful selection of fractal dimension algorithm is required for specific applications.

Line length: an efficient feature for seizure onset detection

A signal feature with low computational burden is presented as an efficient tool for seizure onset detection. The feature was evaluated over a total of. 1,215 hours of intracranial EEG signal from 10 patients. Results confirmed this feature as being useful for seizure onset detection yielding an average delay of 4.1 seconds, 0.051 false positives per hour, and one false negative on a subclinical seizure out of 111 seizures analyzed of which 23 were subclinical.

Detection of seizure precursors from depth-EEG using a sign periodogram transform

Brief bursts of focal, low amplitude rhythmic activity have been observed on depth electroencephalogram (EEG) in the minutes before electrographic onset of seizures in human mesial temporal lobe epilepsy. We have found these periods to contain discrete, individualized synchronized activity in patient-specific frequency bands ranging from 20 to 40 Hz. We present a method for detecting and displaying these events using a periodogram of the sign-limited temporal derivative of the EEG signal, denoted joint sign periodogram event characterization transform (JSPECT). When applied to continuous 2-6 day depth-EEG recordings from ten patients with temporal lobe epilepsy, JSPECT demonstrated that these patient-specific EEG events reliably occurred 5-80 s prior to electrical onset of seizures in five patients with focal, unilateral seizure onsets. JSPECT did not reveal this type of activity prior to seizures in five other patients with bilateral, extratemporal or more diffuse seizure onsets on EEG. Patient-specific, localized rhythmic events may play an important role in seizure generation in temporal lobe epilepsy. The JSPECT method efficiently detects these events, and may be useful as part of an automated system for predicting electrical seizure onset in appropriate patients.

A multi-feature and multi-channel univariate selection process for seizure prediction

OBJECTIVE To develop a prospective method for optimizing seizure prediction, given an array of implanted electrodes and a set of candidate quantitative features computed at each contact location. METHODS The method employs a genetic-based selection process, and then tunes a probabilistic neural network classifier to predict seizures within a 10 min prediction horizon. Initial seizure and interictal data were used for training, and the remaining IEEG data were used for testing. The method continues to train and learn over time. RESULTS Validation of these results over two workshop patients demonstrated a sensitivity of 100%, and 1.1 false positives per hour for Patient E, using a 2.4s block predictor, and a failure of the method on Patient B. CONCLUSIONS This study demonstrates a prospective, exploratory implementation of a seizure prediction method designed to adapt to individual patients with a wide variety of pre-ictal patterns, implanted electrodes and seizure types. Its current performance is limited likely by the small number of input channels and quantitative features employed in this study, and segmentation of the data set into training and testing sets rather than using all continuous data available. SIGNIFICANCE This technique theoretically has the potential to address the challenge presented by the heterogeneity of EEG patterns seen in medication-resistant epilepsy. A more comprehensive implementation utilizing all electrode sites, a broader feature library, and automated multi-feature fusion will be required to fully judge the methods potential for predicting seizures.

A comparison of fractal dimension algorithms using synthetic and experimental data

The fractal dimension (FD) of a waveform represents a powerful tool for transient detection. In particular, in analysis of electroencephalograms (EEG) and electrocardiograms (EGG), this feature has been used to identify and distinguish specific states of physiologic function. A variety of algorithms are available for the computation of FD. In this study, the most common methods of estimating the FD of biomedical signals are analyzed and compared. The analysis is performed over both synthetic data and intracranial EEG (IEEG) data recorded during pre-surgical evaluation of individuals with epileptic seizures. The advantages and drawbacks of each technique are highlighted. The effects of window size, number of overlapping points, and signal to noise ratio (SNR) are evaluated for each method. This study demonstrates that a careful selection of FD algorithm is required for specific applications.

Fractal dimension characterizes seizure onset in epileptic patients

We present a quantitative method for identifying the onset of epileptic seizures in the intracranial electroencephalogram (IEEG), a process which is usually done by expert visual inspection, often with variable results. We performed a fractal dimension (FD) analysis on IEEG recordings obtained from implanted depth and strip electrodes in patients with refractory mesial temporal lobe epilepsy (MTLE) during evaluation for epilepsy surgery. Results demonstrate a reproducible and quantifiable pattern that clearly discriminates the ictal (seizure) period from the pre-ictal (pre-seizure) period. This technique provides an efficient method for IEEG complexity characterization, which may be implemented in real time. Additionally, large volumes of IEEG data can be analyzed through compact records of FD values, achieving data compression on the order of one hundred fold. This technique is promising as a computational tool for determination of electrographic seizure onset in clinical applications.

Feature parameter optimization for seizure detection/prediction

When dealing with seizure detection/prediction problems, there are three main performance metrics that must be optimized: false positive rate, false negative rate, detection delay or, if the problem is seizure prediction, it is desirable to obtain the greatest prediction time achievable. Tuning specific extracted features to individual patients can lead to improved results. The processing window length is also an important parameter whose optimization may significantly affect performance. In this study we propose an approach for selecting the window length for the particular detection/prediction problem. This approach is applicable to other feature parameters suitable for tuning or optimization.

Comparison of line length feature before and after brain electrical stimulation in epileptic patients

This study aims to determine whether there are any statistically significant effects in the intracranial EEG signal due to brain electrical stimulation that can be quantified by comparing the line length value computed in windows positioned before and after stimulated abnormal events versus windows before and after non-stimulated abnormal events. The line length feature has been previously demonstrated to preserve waveform dimensionality changes as the ones estimated by Katzs fractal dimension and is a measure sensitive to variations in signal amplitude and frequency, equivalent in some ways to Teagers energy. Brief stimulation bursts of 200 Hz were delivered in response to some detections of abnormal electrographic activity. A total of 35 epileptic patients were analyzed including 15,938 electrographic events, of which 4,584 were electrically stimulated events. The ratio and difference of the post-stimulation versus the pre-stimulation line length values were computed as comparison measures. The average line length ratios in stimulated events versus those in non-stimulated events were lower in 23 out of 35 patients, suggesting that stimulation may have had an effect on electrographic activity. Statistical analysis based on a permutation test indicated the probability of finding this difference by random chance was 5.21%, further suggesting that the line length ratio differences are most likely due to the stimulation effects on the brain that manifest in the electrographic activity.

High starting torque for AC SCR controller

A control strategy is proposed for AC thyristor controllers. The main feature of the proposed technique is that motors can start with high torque, while using an economical design. This allows the use of AC thyristor controllers for a wide range of applications, where they have not been used before. The method employs the AC thyristor controller as a discrete frequency inverter that increases the frequency until the frequency of the line is reached (60 Hz in this case). These discrete frequencies produced by the control are sub-multiples of line frequency. They are generated by omission or inclusion of line frequency half cycles. Voltage sequences are predefined for each sub-multiple of line frequency. In order to obtain positive torque for all these frequencies, the system is unbalanced. The proposed control is simulated (using EMTP), and built. Comparisons are done with the traditional AC thyristor starter. Results show reduced RMS starting current, increased starting torque and the possibility to operate at low speeds.