Information Structuring and Symbolic Representation for Analysis of Resuscitation Data

Data from resuscitation episodes offers a multitude of problems to study: shock outcome prediction, automated pulse detection, clinical state transitions,... For each of these problems ECG recordings has to be extracted from large data sets. The criteria for extraction differs from problem to problem. We present a method for information structuring of rhythm, therapy and noise annotations and show how it is possible to define extraction criteria for both the outcome prediction and the pulse detection problem preparing further signal analysis. Furthermore we present a method for making symbolic representations of resuscitation data which serves as a starting point to study the complexity of patient episodes. We demonstrate how this representation technique is applied in Markov and intensity modelling of clinical state transitions during resuscitation