Much of the research in my lab focuses on using signal processing techniques and other computational methods to model heart and brain activity. The electrical activation sequence of the heart is most interesting during episodes of abnormal activity, for instance during an arrhythmia (or when the heart beats with an irregular rhythm). More specifically, my project focuses on atrial fibrillation (AF), a condition during which the heart’s atria beat chaotically resulting in poor blood flow to the body.
During an episode of paroxysmal atrial fibrillation, the faulty electrical signals and irregular heartbeat will begin and then stop on its own, so many people with paroxysmal AF will rapidly alternate between sequences of AF rhythm and normal sinus rhythm. However, upon analyzing electrocardiogram data of various cases of AF, I hypothesize that the normal sinus rhythm of someone who had recently experienced a bout of AF or who may be at risk of AF has different properties than the normal sinus rhythm of someone without AF. Although someone with paroxysmal AF will eventually return to and maintain a normal rhythm, this “normal” rhythm is significantly different than the normal rhythm of a non-AF person.
One of these properties that appears to differentiate the normal sinus rhythm of an AF patient versus the normal sinus rhythm of a non-AF patient is heart rate variability, or the variation in the beat-to-beat time interval. Overall, the main objective of my project is to determine a robust measure of heart rate variability and then use logistic regression to model the probability that a person has had or is at risk of an episode of AF. This model could be a potentially powerful diagnostic tool because, unlike the majority of methods that rely on the actual presence of arrhythmia, the algorithm will be able to use information from normal sinus rhythm to predict the possible onset of AF.
