Heartbeat dynamics is a complex system. To charac- terize it properly, advanced nonlinear signal-processing methods are needed. In this context, recent developments on reconstructible signals and multiscale information con- tent have led to the Microcanonical Multifractal Formal- ism (MMF). The MMF provides signal-analysis techniques particularly suited to heartbeat dynamics. In particu- lar, electrocardiogram signals and electric potential in the endocardium allows detecting slow-changing transitions. Detecting regime transition could be used for early warn- ing and treatment of cardiac arrhythmias. In this context, we present an application to the case of Atrial Fibrillation in which we detect distinctive parameters for the transition matrix