Characterization and classification of cardiac dynamics on the basis of measured time series (e.g. electrocardiogram, ECG) is crucial for distinguishing physiological from pathological states with potential applications for diagnosis and risk assessment. Since the heart is an extended system, its spatiotemporal electro-mechanical activities in combination with additional control loops of the cardiovascular system form the basis of most observed signals and any (nonlinear) response to perturbations. In particular, complex spatiotemporal dynamics in terms of rotating action potential waves called spiral waves and their subsequent breakup into additional spiral waves has a strong impact on heart rate variability and may cause many cardiac disorders.
Much progress has been made on the microscopic dynamics underlying cardiac rhythm (i.e., ion channel dynamics) and intermediate-scale dynamics (i.e., individual spiral waves). However, characterization of the macroscopic dynamics, including the global behavior of multiple spiral waves, underlying molecular or genetic mechanisms, and its implications for clinically relevant signals including the ECG remain largely elusive.
Therefore, we determine characteristic features of measured signals that are suitable for classification using state of the art machine learning methods, and we investigate the corresponding underlying spatiotemporal dynamics. The development and application of novel analysis and classification methods require an interdisciplinary and collaborative research environment, integrating data acquisition, analysis and interpretation within the biomedical context.