The influence of lead selection on the correlation dimension for quantification of the chaotic attractor underlying polymorphic tachyarrhythmias

Currently the correlation dimension is frequently used to quantify the non-linear properties of the electrocardiogram recorded in rest or during various tachyarrhythmias. In this study the influence of lead selection and embedding dimension on the correlation dimension was evaluated. In nine dogs presented with Torsade de Pointes tachyarrhythmias the correlation dimension was determined using the commonly applied method of Grassberger and Procaccis, for the embedding dimension 3, 4, 5 and 8. Between dogs the average correlation dimension varied from 1.37-2.12, 1.42-2.39, 1.45-2.63, and 1.55-3.03 respectively. Within dogs the difference between those two leads with the minimal and maximal dimension varied from 0.17-0.67, 0.19-0.95, 0.24-1.00 and 0.32-1.14. The study suggests that the correlation dimension should not be used to compare the non-linearity between patients but should be limited to longitudinal comparison in one patient.