Robert D. Throne

Scatter Diagram Analysis: A New Technique for Discriminating Ventricular Tachyarrhythmias

With the increasing flexibility allowed by implantable cardioverter defibrillators that use tiered therapy, it is important to match the therapy with the arrhythmia. In this article we present scatter diagram analysis, a new computationally efficient two‐channel algorithm for distinguishing monomorphic ventricular tachycardia (VT) from polymorphic ventricular tachycardia and ventricular fibrillation (VF). Scatter diagram analysis plots the amplitude from one channel versus the amplitude from another channel on a graph with a 15 × 15 grid. The fraction (percentage) of the 225 grid blocks occupied by at least one sample point is then determined. We found that monomorphic VT traces nearly the same path in space and occupies a smaller percentage of the graph than a nonregular rhythm such as polymorphic VT or VF. Scatter diagram analysis was tested on 27 patients undergoing intraoperative implantable cardioverter defibrillator testing. Passages of 4.096 seconds were obtained from rate (bipolar epicardial) and morphology (patch) leads, and digitized at 125 Hz. Scatter diagram analysis distinguished 13 episodes of monomorphic VT (28.6%± 4.0%) from 27 episodes of polymorphic VT or VE (48.0%± 8.2%) with P < 0.0005. There was overlap in only one monomorphic VT episode and one polymorphic VT or VF episode.

A comparison of higher-order generalized eigensystem techniques and tikhonov regularization for the inverse problem of electrocardiography *

In a recent series of papers we proposed a new class of methods, the generalized eigensystem (GES) methods, for solving the inverse problem of electrocardiography. In this paper, we compare zero, first, and second order regularized GES methods to zero, first, and second order Tikhonov methods. Both optimal results and results from parameter estimation techniques are compared in terms of relative error and accuracy of epicardial potential maps. Results from higher order regularization depend heavily on the exact form of the regularization operator, and operators generated by finite element techniques give the most accurate and consistent results. In the optimal parameter case, the GES techniques produce smaller average relative errors than the Tikhonov techniques. However, as the regularization order increases, the difference in average relative errors between the two techniques becomes less pronounced. We introduce the minimum distance to the origin (MDO) technique to choose the number of expansion modes ...

Performance of generalized eigensystem and truncated singular value decomposition methods for the inverse problem of electrocardiography

Singular Value Decomposition (SVD) and Generalized Eigensystem (GES) inverse techniques are compared for their ability to solve the inverse problem of electrocardiography. In the inverse problem of electrocardiography, electrical potential data for numerous locations on the body (torso) surface is used to infer the electrical potentials on the heart surface, while the governing equation and material properties are assumed known. This paper addresses two areas. First, the previously observed improved performance of GES compared to SVD is explained in terms of the unique nature of the GES vectors. Second, epicardial data from six in-vitro rabbit heart experiments are used to project body surface data for six different geometries, and inverse solutions are computed both with and without added noise. For concentric geometries, GES outperformed SVD in all instances. For eccentric heart/body geometries, GES outperformed SVD when the inverse errors themselves were small. In all cases, GES was less sensitive to a...

Detecting ventricular fibrillation using efficient techniques for computing a normalized autocorrelation

With the introduction of the multifunction implantable pacemaker/cardioverter/defibrillator, it is increasingly important to detect and identify arrhythmias automatically. Detection of ventricular fibrillation by analysis of the autocorrelation function is widely used on surface lead ECG analysis, but due to the computational demand is not practical for use in an implantable defibrillator. In this paper, results using three computationally efficient algorithms for estimating the normalized autocorrelation are compared with the true normalized autocorrelation for discriminating polymorphic ventricular tachycardia/ventricular fibrillation (PMVT/VF) from monomorphic ventricular tachycardia (MVT) using signals available to an implantable defibrillator.

Estimation of tool/chip interface temperatures for on-line tool monitoring: an inverse problem approach

We examine a steady inverse heat transfer problem that arises in online machine tool monitoring: identifying tool/chip interface temperatures from remote sensor measurements. The matrix equations relating the sensor temperatures and sensor fluxes to the prediction surface (tool/chip interface) temperatures are obtained by finite element methods. Truncated singular value decomposition, a standard inverse technique, is used as a baseline for comparing the inverse solutions. We also develop a new set of inverse approaches, vector projection inverse methods, specifically for this problem. Inverse solutions are computed with all methods for two temperature profiles and various noise levels. Because of the extreme ill-conditioning of the problem, only two coefficients can be obtained reliably for all of the inverse approaches examined. Truncated singular value decomposition does not perform well, but two of the new methods are robust and give reasonable accuracy. Combining data from temperature and flux sensors (data fusion) is far more effective than using temperature sensors alone, and with data fusion the inverse can be computed robustly with information from only four sensor locations.

A comparison of generalized eigensystem, truncated singular value decomposition, and tikhonov regularization for the steady inverse heat conduction problem*

We present new techniques for solving inverse boundary value problems in steady heat conduction. These new Generalized Eigensystem techniques are vector expansion methods which have previously been used in inverse electrocardiography applications. We compare the Generalized Eigensystem techniques to truncated singular value decomposition and Tikhonov regularization on two two-dimensional test geometries and four temperature/flux patterns. One of the Generalized Eigensystem methods (GESL) substantially outperforms the other techniques studied on the majority of the test cases, with inverse errors up to 20 times smaller than other approaches. In addition, even when the number of sensors on the boundary is reduced, GES L was still comparable to or superior to the other techniques with a full sensor set.

Improved parameter choice methods and temporal filtering for the generalized eigensystem method applied to the inverse problem of electrocardiography

We have previously proposed the generalized eigensystem (GES) method as a modal expansion method for estimating electrical potentials on the heart outer surface from measurements of electrical potentials on the body surface. In this paper, we present an alternative formulation of GES more like that of classical Tikhonov regularization where a single continuous parameter needs to be chosen. We then compare this formulation of GES with zero order Tikhonov regularization on data collected from a swine experiment with the swine heart paced from six different sites. Of the two algorithms used to estimate the regularization parameters, the composite residual and smoothing operator (CRESO) generally outperforms the generalized cross validation (GCV) for both GES and Tikhonov. Although the inverse problems are solved at each time instant independently, we also incorporate temporal information by moving average filtering the estimates, and this is more effective for the GES methods than for Tikhonov. In general, the epicardial estimates using the GES method are more accurate than the estimates generated using zero order Tikhonov regularization.

A Comparison Of Five Methods For Construction Of Regularization Operators For Higher-Order Tikhonov Regularization

Tikhonov regularization is widely used in, solving the ill-conditioned inverse problem of electrocardiograp h y. Zero- ord er, first- order, and second- order regularization have all been tested, with varying results for the higher-order operators. This study compares five methods for generating the operators. Results from higher order regularization depend heavily on, the exact form of the regularization operator, and of the operators studied here those generated by finite element techniques (weighted residual methods) appear io give the most accurate and consistent results.

Automating the selection of expansion modes using the principal components of time method for solving the inverse problem of electrocardiography

A common approach for estimating epicardial potentials from measured body sutface potentials is to solve the problem using some type of regularization at each point in time independently. Greensite and Huiskamp huve previously proposed a solution based on expanding the solution in the principle time components of the data. However; their proposed method did not indicate how many modes to use in this expansion, or indicate how to automate it. In this paper; we examined an automated solution to this problem. This automated procedure was applied to data obtained during a study on swine. For first order Ekhonov regularization, the new method produced an averuge relative error of 0.56 over all six depolarization sequences, compared with 0.61 for the traditional approach. While the new automated method is not necessarily optimal, it does consistently produce smaller relative errors than the traditional approach.

An open environment for reconstructing 2D images into 3D finite element models

Eventual clinical utilization of various proposed techniques for determining the epicardial potentials based on measurements on the body surface requires the rapid development of accurate human torso models. These models are generally constructed using data obtained via competed tomography (CT) or magnetic resonance imaging (MRI). Automatic and/or manual segmentation of the anatomical items in the individual slices is typically done as a preprocessing step before surface reconstruction. The authors have developed a group of interactive, menu-driven routines to assist in processing these CT or MRI slices for input into a commercial finite element package for subsequent finite element modeling of the torso.