engleski

Nonlinear Information Processing in Extracting Predictive Attributes of an ECG Signal

We studied three target groups of ECG recordings. The first group are the records 5 minutes before of paroxysms of atrial fibrillation (PAF). The second group are records far from PAF. The last group consists of records from healthy heart dynamics. The goal was to extract system dimensions as a possible predictive attribute. Specifically, we were looking for dimension changes (from minute to minute) in ECG recordings. We applied surrogate data method to check the possible level of determinism in data. The results clearly exhibited the fluctuations in dimensions in the last 5 minutes in most of the examples from all target groups. It was not possible to create general dimensional differences from one target group to the other, but it was not feasible to extract many smaller subgroups in every target group that have a common dimensional characteristics, such as the limits on dimensions at the particular observed time. The relative dimension change from minute to minute, as a possible predictive attribute, shows some subgroup of rules important for partial understanding of apperance of PAF.

Nonlinear systems ; Fibrillation ; Predictors ; Dimension

nije evidentirano