Using support vector machines for atrial fibrillation screening

Atrial fibrillation is the most common type of arrhythmia in the world. Therefore, the development of suitable techniques for detection of atrial fibrillation is of paramount importance, particularly for ambulatory screening of potential patients. On the other hand, the current availability of wearable devices capable of monitoring biometric signals opens the door for such ambulatory screening to be affordably performed at a large scale. One of the main challenges in this context is the need for analyzing huge amounts of data. Computational intelligence techniques are good candidates to carry out this kind of analysis. This paper proposes the use of Support Vector Machines for heart rhythm analysis, aiming at discriminating between atrial fibrillation and sinus rhythm. Heart rate information is obtained from photoplethysmography sensors embedded in wearable devices. Experimental results obtained from real patients are presented and discussed, pointing to the suitability of the proposed solution for affordable atrial fibrillation screening.