| Abstract [eng] |
The current paradigm provides a scientific basis to assume that temporal atrial fibrillation (AF) episode pattern is associated to the risk of thrombus formation. That is, the risk may potentially be higher when AF episodes are aggregated in time since the flow velocity in the left atrial appendage decreases during AF. While information about AF patterns is lacking, the emerging non-invasive technologies for long-term monitoring are expected to fill in this gap in knowledge. However, in order to properly tackle this problem, approaches to characterizing the variety of types of patterns are needed. In this doctoral thesis, a clinically relevant scientific-technological problem of characterization of temporal distribution of paroxysmal AF episodes is covered. For this purpose, three different approaches to characterize AF patterns have been proposed and investigated. One of them is based on the statistical distribution analysis which rests on the assumption that episodes are statistically independent. This assumption may be questioned since AF episodes tend to cluster. Another approach is to use parameters from the AF pattern model which is based on an alternating, bivariate Hawkes process. The model-based parameters provide information on AF episode clustering and rhythm dominance. The final approach is to use various parameters, i.e., AF burden, well-known parameter in AF studies, provides information only about the time spent in AF; meanwhile, aggregation provides information on the temporal distribution of AF episodes, and Gini coefficient provides information on the episode duration inequality. A combination of model-based and parameter-based characterization offers a solution to analyze different types of AF patterns. |
