New developments in the automatic analysis of the surface ECG: the case of atrial fibrillation.

T his article discusses current approaches to the analysis of digital electrocardiographic data, with special reference to the detection and classification of episodes of atrial fibrillation (AF). AF is the most common cardiac arrhythmia encountered in clinical practice. It is a supraventricular tachyarrhythmia characterised by uncoordinated atrial activation, with resulting deterioration of atrial mechanical functionality. AF accounts for approximately one third of hospitalisations for cardiac rhythm disturbance, affecting 4.5 million people in the EU and up to 2.3 million people in the USA. However, the public health burden of AF could increase dramatically over the next 50 years in parallel with the ageing population, since the prevalence and the incidence of AF are known to increase with age. The guidelines drawn up by the American College of Cardiology, American Heart Association, and the European Society of Cardiology include a classification scheme for AF that is unrelated to an immediate, reversible cause (which would mean that the AF is not likely to recur or to have prognostic significance after the disappearance of the acute aetiology). The scheme specifies four categories: first detected episode (provides no knowledge of previous events and uncertainty about duration); paroxysmal (recurrent, but can terminate spontaneously in less than 7 days, and commonly less than 48 hours); persistent (recurrent, lasting more than 7 days, and typically requires cardioversion to terminate); and permanent (fails to terminate after cardioversion, or relapses less than 24 hours after termination, or no cardioversion attempted). Note that when a patient has had 2 or more episodes after the clinical detection of the first episode of AF, the arrhythmia is considered to be recurrent. Thus, the guidelines focus on the classification and treatment of the arrhythmia based on its duration (paroxysmal, persistent and permanent) rather than on its aetiology. AF is associated with substantial mortality and morbidity due to thromboembolism, heart failure, and impaired cognitive function. In particular, persistent or permanent AF entails a serious risk of thromboembolism, as a result of thrombus formation within the atria, which can cause stroke or other thromboembolic events. Paroxysmal AF accounts for 35 to 66% of total cases and may deteriorate into persistent or permanent AF. Considering that the ideal therapeutic goal for AF is to achieve and maintain sinus rhythm, the use of aggressive procedural techniques should be avoided in the case of paroxysmal episodes. Thus, the discrimination between paroxysmal and persistent or permanent AF, and the prediction of paroxysmal AF termination, can be invaluable in order to avoid useless therapeutic interventions, reduce the assoNew Developments in the Automatic Analysis of the Surface ECG: The Case of Atrial Fibrillation