Patrick Celka

A computer-aided detection of EEG seizures in infants: a singular-spectrum approach and performance comparison

Presents a scalp electroencephalogram (EEG) seizure detection scheme based on singular spectrum analysis (SSA) and Rissanen minimum description length (MDL) model-order selection (SSA-MDL). Preprocessing of the signals allows for the drastic reduction of the number of false alarms. Statistical performance comparison with seizure detection schemes of Gotman et al. (1997) and Liu et al. (1992) is performed on both synthetic data and real EEG seizures. Monte Carlo simulations based on synthetic infant EEG seizure data reveals some detection drawbacks on a large variety of seizure waveforms. Detection using both Monte Carlo and four real infant scalp EEG signals shows the superiority of the SSA-MDL method with an average good detection rate of >93% and false detection rate <4%.

Preprocessing and time-frequency analysis of newborn EEG seizures

Neurological disease or dysfunction in newborn infants is often first manifested by seizures. Prolonged seizures can result in impaired neurodevelopment or even death. In adults, the clinical signs of seizures are well defined and easily recognized. In newborns, however, the clinical signs are subtle and may be absent or easily missed without constant close observation. This article describes the use of adaptive signal processing techniques for removing artifacts from newborn electroencephalogram (EEG) signals. Three adaptive algorithms have been designed in the context of EEG signals. This preprocessing is necessary before attempting a fine time-frequency analysis of EEG rhythmical activities, such as electrical seizures, corrupted by high amplitude signals. After an overview of newborn EEG signals, the authors describe the data acquisition set-up. They then introduce the basic physiological concepts related to normal and abnormal newborn EEGs and discuss the three adaptive algorithms for artifact removal. They also present time-frequency representations (TFRs) of seizure signals and discuss the estimation and modeling of the instantaneous frequency related to the main ridge of the TFR.

Stochastic analysis of gradient adaptive identification of nonlinear systems with memory for Gaussian data and noisy input and output measurements

This paper investigates the statistical behavior of two gradient search adaptive algorithms for identifying an unknown nonlinear system comprised of a discrete-time linear system H followed by a zero-memory nonlinearity g(/spl middot/). The input and output of the unknown system are corrupted by additive independent noises. Gaussian models are used for all inputs. Two competing adaptation schemes are analyzed. The first is a sequential adaptation scheme where the LMS algorithm is first used to estimate the linear portion of the unknown system. The LMS algorithm is able to identify the linear portion of the unknown system to within a scale factor. The weights are then frozen at the end of the first adaptation phase. Recursions are derived for the mean and fluctuation behavior of the LMS algorithm, which are in excellent agreement with Monte Carlo simulations. When the nonlinearity is modeled by a scaled error function, the second part of the sequential gradient identification scheme is shown to correctly learn the scale factor and the error function scale factor. Mean recursions for the scale factors show good agreement with Monte Carlo simulations. For slow learning, the stationary points of the gradient algorithm closely agree with the stationary points of the theoretical recursions. The second adaptive scheme simultaneously learns both the linear and nonlinear portions of the unknown channel. The mean recursions for the linear and nonlinear portions show good agreement with Monte Carlo simulations for slow learning. The stationary points of the gradient algorithm also agree with the stationary points of the theoretical recursions.

Nonlinear nonstationary Wiener model of infant EEG seizures

This paper presents the estimation of a nonstationary nonlinear model of seizures in infants based on parallel Wiener structures. The model comprises two parts and is partly derived from the Roessgen et al. seizure model. The first part consists of a nonlinear Wiener model of the pure background activity, and the second part in a nonlinear Wiener model of the pure seizure activity with a time-varying deterministic input signal. The two parts are then combined in a parallel structure. The Wiener model consists of an autoregressive moving average filter followed by a nonlinear shaping function to take into account the non-Gaussian statistical behavior of the data. Model estimation was performed on 64 infants of whom four showed signs of clinical and electrical seizures. Model validation is performed using time-frequency-based entropy distance and shows an averaged improvement of 50% in modeling performance compared with the Roessgen model.

Wrist-located pulse detection using IR signals, activity and nonlinear artifact cancellation

We present a new integrated device for monitoring heart rate at the wrist using an optical measure. Motion robustness is obtained by using accurate motion reference signals of 3D low noise accelerometers together with dual channel optical sensing. Nonlinear modelling allows to remove the motion contributions in the optical signals and the spatial diversity of the sensors is used to remove reciprocal contributions in the two channels. Finally a statistical estimation, based on physiological properties of the heart, gives a robust estimation of the heart rate. Qualitative and quantitative performance evaluation of the performances on real signals clearly show that the proposed system gives an accurate estimation of the heart rate, even under intense physical activity.

Stochastic gradient identification of polynomial Wiener systems: analysis and application

This paper presents analytical, numerical, and experimental results for a stochastic gradient adaptive scheme that identifies a polynomial-type nonlinear system with memory for noisy output observations. The analysis includes the computation of the stationary points, the mean square error surface, and the stability regions of the algorithm for Gaussian data. Convergence of the mean is studied using L/sub 2/ and Euclidian norms. Monte Carlo simulations confirm the theoretical predictions that show a small sensitivity to the observation noise. An application is presented for the identification of a nonlinear time-delayed feedback system.

Analysis of stochastic gradient identification of Wiener-Hammerstein systems for nonlinearities with Hermite polynomial expansions

This paper investigates the statistical behavior of a sequential adaptive gradient search algorithm for identifying an unknown Wiener-Hammerstein (1958) system (WHS) with Gaussian inputs. The WHS nonlinearity is assumed to be expandable in a series of orthogonal Hermite polynomials. The sequential procedure uses (1) a gradient search for the unknown coefficients of the Hermite polynomials, (2) an LMS adaptive filter to partially identify the input and output linear filters of the WHS, and (3) the higher order terms in the Hermite expansion to identify each of the linear filters. The third step requires the iterative solution of a set of coupled nonlinear equations in the linear filter coefficients. An alternative scheme is presented if the two filters are known a priori to be exponentially shaped. The mean behavior of the various gradient recursions are analyzed using small step-size approximations (slow learning) and yield very good agreement with Monte Carlo simulations. Several examples demonstrate that the scheme provides good estimates of the WHS parameters for the cases studied.

Observer of autonomic cardiac outflow based on blind source separation of ECG parameters

We present a novel method which provides an observer of the autonomic cardiac outflow using heartbeat intervals (RR) and QT intervals. The model of the observer is inferred from qualitative physiological knowledge. It consists in a problem of blind source separation of noisy mixtures which is resolved by a simple and robust algorithm. The robustness of the algorithm has been assessed by numerical simulations in adverse noisy environments. In clinical applications, we have validated the observer on subjects exposed to experimental conditions known to elicit sympathetic or parasympathetic response.

Observer of the human cardiac sympathetic nerve activity using noncausal blind source separation

We present a novel method for the blind reconstruction of the cardiac sympathetic nerve activity (CSNA) in the low-frequency (LF) band (0.04-0.15 Hz) using only heart rate and arterial blood pressure. The originality of the method consists in the application of blind source separation techniques to obtain an observer of CSNA. We show how this observer can be deduced from a linear model of the cardiovascular system by introduction of the fundamental assumptions about the independence of the cardiac sympathetic an parasympathetic outflow. In cardiovascular applications, the reliability of the observer has been assessed by verification of the fundamental assumption for the given data. A primer qualitative validation has been performed using the muscle sympathetic nerve activity as an indirect indicator of CSNA. Very satisfying and promising results have been obtained. Moreover, we have performed quantitative validations of the observer in various experimental conditions known to elicit selectively cardiac sympathetic or parasympathetic response. The experimental conditions include a supine-to-60/spl deg/ tilt test, indirect sympathetic stimulation/inhibition by medication, and sympathetic stimulation by isometric handgrip. We show that the observer allows to highlight changing levels of the cardiac sympathetic activity in the LF band in all these experimental conditions.

Analysis of stochastic gradient tracking of time-varying polynomial Wiener systems

This paper presents analytical and Monte Carlo results for a stochastic gradient adaptive scheme that tracks a time-varying polynomial Wiener (1958) system [i.e., a linear time-invariant (LTI) filter with memory followed by a time-varying memoryless polynomial nonlinearity]. The adaptive scheme consists of two phases: (1) estimation of the LTI memory using the LMS algorithm and (2) tracking the time-varying polynomial-type nonlinearity using a second coupled gradient search for the polynomial coefficients. The time-varying polynomial nonlinearity causes a time-varying scaling for the optimum Wiener filter for Phase 1. These time variations are removed for Phase 2 using a novel coupling scheme to Phase 1. The analysis for Gaussian data includes recursions for the mean behavior of the LMS algorithm for estimating and tracking the optimum Wiener filter for Phase 1 for several different time-varying polynomial nonlinearities and recursions for the mean behavior of the stochastic gradient algorithm for Phase 2. The polynomial coefficients are shown to be accurately tracked. Monte Carlo simulations confirm the theoretical predictions and support the underlying statistical assumptions.