Data mining for detecting disturbances in heart rhythm

In this paper a framework for an objective and interactive grading system of disturbances in heart rhythm is presented. The objects included in the study are transgenic mice which suffer from cardiac arrhythmias and their wild-type siblings. For all these mice long-time ECG recordings are available. The RR interval length are utilised to deduce statistical features on short-time heartbeat patterns. We demonstrate that these features are biologically relevant to classify the heartbeat patterns by a K-means clustering approach. Additionally, an extension of K-means is proposed, which consider user-defined constraints. The results of constraint-based clustering methods enable significant mid- and long-time analysis studies.