E.J. Berbari

The analysis of ventricular late potentials using orthogonal recordings

The use of orthogonal lead recordings to characterize signal-averaged late potentials is discussed. A theoretical analysis is made of the performances of individual orthogonal leads and their vector magnitude. Some limitations of the vector magnitude for characterizing late potentials are developed in relation to the distribution of signal and noise among its component orthogonal leads. Starting from the basic statistical assumptions required for signal averaging, an analysis of late potentials was simulated using ideal signal and noise functions. The results from theory and practice show an ability to predict the relative performances of signal-averaged orthogonal leads and their vector magnitude under typical recording conditions. This is illustrated with some observations from patient data. The conclusion of this investigation is that the vector magnitude waveform is not always able to reflect accurately the original information available in the signal-averaged orthogonal leads.<<ETX>>

Orthogonal Surface Lead Recordings of His-Purkinje Activity: Comparison of Actual and Simulated Waveforms

Noninvasive His-Purkinje system (HPS) recordings can be obtained using high amplification and signal averaging. However, this technique lacks the firm underpinnings necessary for its establishment as a useful clinical tool for the study of the electrical activity of the heart. Previous work has been incomplete resulting in fragmented knowledge of the body surface HPS activity and the nature of the HPS as an electrical source.

Use of time-frequency representations to enhance the signal-to-noise ratio of the signal-averaged ECG

The signal-averaged electrocardiogram (SAECG) increases the signal-to-noise ratio of the electrocardiogram to recover low-level signals, such as ventricular late potentials. The authors present some preliminary results from the use of an optimal filter which can increase the signal-to-noise ratio of the SAECG more efficiently than signal averaging. The setting for the filter is the time-frequency plane. The optimal filter was tested on eight recordings, comparing signal averages performed on a 0.3- mu V root-mean-square (RMS) noise endpoint with 64 beat averages before and after filtering. It was concluded that optimal filtering to enhance the SNR of the SAECG in the setting of the time-frequency plane is feasible.<<ETX>>

A cardiac mapping system for identifying late potentials: correlation with signal averaged surface recordings

Cardiac late potentials, measured with the signal-averaged ECG, have been strongly correlated with potentially lethal ventricular arrhythmias. A key factor in identifying and quantifying these late potentials is the determination of the end of ventricular depolarization. The goal of this study was to identify the latest moment of ventricular depolarization by comparing a signal-averaged ECG with multielectrode cardiac mapping of infarct regions. The four-day-old, canine myocardial infarction model was used. In 10 of 13 dogs studied, a concordance in timing between the epicardial array and signal-averaged XYZ surface leads was found. In three dogs the lack of concordance is attributed to a poor signal-to-noise ratio of the surface-signal average and its distortion by high-pass filtering, which often eliminates enough signal energy to preclude identification. Further improvements in low-level signal recovery and postaveraging analysis are needed to ensure that cardiac late potentials are fully identified from body-surface recordings.<<ETX>>

Identifying uncertainty in epicardial activation maps

Epicardial activation maps in surviving infarct regions are derived by: (1) choosing a specific activation time at each recording site and (2) interpolating the data so that contours can be drawn representing the activation wavefronts. Concentrating on the latter problem, the spatial autocorrelation function (SAF) was calculated to evaluate the smoothness or regional correlation of activation times obtained at neighboring recording sites. Low values of the SAF would imply discontinuities (dead tissue) in the underlying data. These represent faults and traditional methods of interpolation cannot be applied as they would assume an activation time over dead tissue. Hence, contour generation must map around these faults. Another approach for defining uncertainty is to examine ambiguous electrograms and choose either the earliest or latest activation times. Two activation maps are created which can be differenced to form an isopachous map. This isopach then depicts the region of uncertainty. These methods have been used in a data set of 17 dogs with four day-old infarcts. The presence and extent of the uncertainty region have a corresponding influence on the visualization of the map and its interpretation.<<ETX>>

Pathophysiological insights into abnormal intra-QRS signals in the high resolution ECG

Presents the idea of abnormal intra-QRS potentials, analogous to conventional late potentials, in the experimental setting of canine myocardial infarction. The link between delayed or slow conduction-evidenced by late potentials-and a reentry mechanism for ventricular tachycardia has been well established. However, the question of whether abnormal intra-QRS potentials may be arrhythmogenic has not been addressed. It is hypothesized that these potentials may reflect disruption and/or delay in ventricular conduction that is potentially arrhythmogenic. If verified, the concept of late potentials might be usefully expanded to include abnormal potentials occurring anywhere within the period of the high resolution QRS complex, as opposed to only the terminal portion.

Performance assessment of optimal filtering of the high resolution electrocardiogram

The high resolution electrocardiogram (HRECG) currently requires ensemble averaging of 200-600 beats to produce a high fidelity signal estimate. An optimal filter in the setting of the time-frequency plane has the potential to enhance the signal-to-noise ratio of the ensemble average significantly. Unfortunately, optimal filters often introduce unacceptable signal distortion. The authors report a performance assessment of the time-frequency plane optimal filtering method. The perspective is of improving overall signal fidelity while reducing noise, with intended superior performance compared to ensemble averaging.<<ETX>>

Contouring of epicardial activation using spatial autocorrelation estimates

The authors examine the procedure involved in producing contour maps of epicardial activation. The key step is to generate a mathematical model of activation. Spectral analysis of heart surface activation reveals a low frequency trend, representing uniform conduction, with disordered conduction concentrated in higher frequency bands. A method of mapping which uses the spatial autocorrelation function to assist in generating the epicardial activation model was compared with traditional, wholly empirical methods. Use of spatial autocorrelation appears to improve the resolution of map generation.<<ETX>>

Use of surface Laplacian to extract local activation from epicardial electrograms

Electrocardiograms were recorded overlying a region of injury on the epicardial surface of a canine heart in a four day infarct model. Electrograms were recorded simultaneously from 121 sites with an electrode array with interelectrode spacing of 1 mm. The surface Laplacian, L, of the electrograms was calculated using the four neighboring sites. Propagation in the border zone was modeled by a planar wave of finite thickness moving with uniform velocity. The source region was simulated with use of the bidomain model for the heart. Computer model results showed that the amplitude of L, diminished as electrode spacing became larger than 0.1 mm. The amplitude also fell as the distance of the wave from the surface increased, decreasing by 40% for a depth of 0.1 mm. The surface Laplacian determined from recorded electrograms showed a sharp deflection with a substantial reduction in the QS wave arising from distant sources. Hence the surface Laplacian may be useful as a spatial filter to extract local activity from epicardial electrograms.

Interpolation of local activation times versus potentials to derive an activation map in infarcted myocardium

Cardiac mapping is a tool used in the analysis of activation wavefront patterns. The highly variable nature of infarcted myocardium may produce significant errors in activation mapping, using traditional mapping techniques. There are two aims to this study. The first aim is to compare activation time detection methods and the second aim is to compare potential and time interpolation. Unipolar electrograms recorded from epicardium of an infarct in the left ventricle of a canine were used to simulate activation patterns over an infarct region. Detected activation times, using (-dV/dt)/sub max/,, and the Laplacian, a spatial derivative, were compared. The Laplacian, unlike the (-dV/dt)/sub max/ could distinguish local from distant events. However, the Laplacian had an unexpected systematic error which was dependent on the direction a velocity of activation. Activation map generation was compared using two different sequences of analysis. Activation time interpolation and potential interpolation. Interpolated activation times could produced precise maps which were independent of propagation velocity. The accuracy of potential interpolation was limited as a function of decreasing conduction velocity.