Adaptive filtering of evoked potentials using higher-order adaptive signal enhancer with genetic-type variable step-size prefilter.

An adaptive signal enhancer based on third-order statistics with a genetic-type, variable step-size prefilter is introduced to recover evoked potentials (EPs). EPs are usually embedded in the ongoing electroencephalogram with a very low signal-to-noise ratio (SNR). As a higher-order statistics technique has a natural tolerance to Gaussian noise, it is applicable for filtering EPs. An adaptive signal enhancer based on third-order statistics was used as the major filter in this study. However, the efficiency of the adaptive signal enhancer was reduced when the total power of uncorrelated noises was large. To improve the performance for EPs under poor SNR, a low-noise signal is required. Therefore a prefilter with a genetic-type, variable step-size algorithm was employed to enhance the SNR of the signal in this study. The fundamental idea of a genetic-type, variable step-size algorithm is that its step-sizes are regularly readjusted to optimum. Therefore this algorithm can be used as a prefilter with different noise levels. Experimental results showed that, for filtering EPs, the proposed scheme is superior to the adaptive signal enhancer with a normalised least mean square algorithm.

Non-invasive assessment of arterial distension waveforms using gradient-based hough transform and power Doppler ultrasound imaging.

The peripheral arterial vessel often appears as an elliptic shape under the constraints of the surrounding tissues. In this study, the gradient-based Hough transform was used to detect the central location of the ellipse and the lumen area of the arterial vessel non-invasively using power Doppler imaging. Sequential ultrasound images were used to construct arterial distension waveforms in both the major- and minor-axis directions for a few cardiac cycles. The common carotid arteries (CCAs) for nine healthy male volunteers (mean age 24 years), in the sitting position, were investigated in vivo. The CCAs (n = 9) had a mean diameter of 5.83mm, and the pulsatile diameter distension was 13.7+/-1.9%. The brachial artery and dorsalis pedis artery for five healthy male volunteers (mean age 26 years), in the supine position, had mean diameters of 4.03mm and 2.83mm and distensions of 16.7+/-4.6% and 15.5+/-5.4%, respectively. The movement of the arterial centre location during the cardiac cycle reflected the asymmetry of the reaction forces produced by the surrounding soft tissues. The present method can obtain the response of vessel distension to pulse pressure, as well as the constrained conditions of the arteries.

Source localization of averaged and single EEG spikes using the electric dipole model.

A scheme for localizing the epileptic focus is proposed. The scheme is on the basis of a model with an electric dipole inside a four-layer inhomogeneous spherical model of head and utilizes a nonlinear programming algorithm applying the gradient projection method. Various initial estimates are used to prove the stability of the implemented dipole model. Fourteen single spike data and the averaged spike data are used to localize the epileptic focus. The results of the usage of the averaged spike data show that the dipole position is compatible with visual inspection of experienced clinical physicians. The results of the usage of the single spike show that 11 of the 14 single spikes have dipole locations near the result of the averaged spike but the estimated dipole moments differ markedly from one another. The localization results of the other three single spikes show that the dipole position is strongly affected by the background EEG. This kind of interference generally causes the eccentricity of the dipole to deviate from its anatomically meaningful value. According to our results, the electric dipole model is concluded to be valuable for the clinical application of localizing epileptic foci.

Evaluation of parametric methods in EEG signal analysis.

In this paper, a well designed database, considering statistical characteristics and including all types of Electroencephalogram (EEG) is built. 900 EEG segments, each with a short interval (1.024 sec) in this database are clustered into eight classes. Three tests of white noise for evaluating the efficiency of autoregressive (AR) and autoregressive-moving average (ARMA) models are proposed. The Akaike information criterion (AIC) is used for determining orders of AR and ARMA models. The AR model requires a higher model order (8.67 on the average) than the ARMA model (6.17 on the average). However, it is found that about 96% of the 900 segments can be efficiently represented by the AR model, and only about 78% of them can be efficiently represented by ARMA model. We therefore conclude that the AR model is preferred for estimating EEG signals.