Assessment of vulnerability to ventricular arrhythmias using magnetocardiographic QRST integral mapping R. Hren 1 , U. Steinhoff 2 , P. Endt 2 , P. Goedde 3 , R. Agrawal 3 , M Oeff 3 , R.L. Lux 4 , and L. Trahms 2 1 Institute of Mathemtics, Physics, and Mechanics, University of Ljubljana, Ljubljana, Slovenia, 2 Physikalisch- Technische Bundesanstalt, Berlin, Germany, 3 Universitaetsklinikum Benjamin Franklin, Berlin, Germany, 4 Nora Eccles Harrison CVRTI, University of Utah, Salt Lake City, USA 1 Introduction It has been shown that regional disparities of ventricular primary repolarization properties can be reflected in body surface distributions of electrocardiographic QRST deflection areas (integrals) [1-4]. To quantitatively assess complexity of spatial features of QRST-integral distributions, several authors (see, e.g., [2,4] and references therein) used orthogonal-expansion methods. In their seminal work, Hubley-Kozey et al. [4] provided evidence that extraction of spatial features of QRST integral maps is suitable for identifying patients at risk of ventricular arrhythmias. Recently, Stroink et al. [5] reported different fragmentation of electric and magnetic ST- T trajectory plots between patients who are at risk of recurrent arrhythmias and those that are not, but found no differences between groups of patients in features of QRST integral maps. In our earlier study [6], we indicated that magnetocardiographic (MCG) QRST integral maps also offer information about the regional electrophysiological properties of the ventricles. We corroborated findings obtained with electrocardiographic QRST integral maps [4] by showing that patients who are at risk of lethal arrhythmias have a markedly higher index of nondipolar content than do healthy subjects. However, nondipolar QRST integral maps can also be found in patients with myocardial infarction (MI) but no clinical arrhythmias [4,7]. The major clinical problem is thus differentiating patients with a history of sustained ventricular tachycardia (VT) and/or ventricular fibrillation (VF) and postinfarction patients who have no history of VT [4]. In this study, we addressed this problem by applying orthogonal-expansion techniques to magnetocardiographic QRST integral maps. 2 Methods 2.1 Patient population and recording techniques Magnetic field maps were recorded during sinus rhythm from 49 high-resolution magnetic leads [8] above the anterior chest in 28 patients (mean age: 61 years) with VT/VF and in 45 patients (mean age: 61) with MI but no VT/VF (MI/non-VT). All MCG recordings were performed in the magnetically shielded room located in the Universitatsklinikum Benjamin Franklin, Berlin. All 49 leads were recorded simultaneously at a sampling rate of 1000 Hz (with 15-bit amplitude resolution) using superconducting quantum inteference device (dcSQUID) first order gradiometers (baseline 7 cm). Analog high pass (RC type) and low pass (Bessel type) filtering was applied to magnetocardiograms at 0.16 and 250 Hz. The UP segment was taken as the zero reference line for all measurements. The MCG signals were processed off-line and averaged over about 80 cardiac cycles. 2.2 Eigenvector analysis In each subject QRST integral maps were calculated and then examined using spatial Karhunen-Loeve (KL) transformation. We calculated the percentage nondipolar content (relative contribution of all KL expansion coefficients beyond the third) as has been described in detail elsewhere [4,6]. Statistical analysis was performed using the Wilcoxon rank sum test since a normal distribution of the values could not be assumed. 3 Results Figures 1 and 2 show QRST integral maps for patients with VT/VF and patients with MI but no VT. We represented the entire pattern space of these QRST integral maps with eigenvectors. The first three eigenvectors had a relatively smooth dipolar distribution, while the higher order (>3) eigenvectors were multipolar with an increasing number of extrema. Mean values of KL transform