An Algorithm for the Detection of the Atrial Fibrillation from the Surface ECG for an Home-Care Evaluation of the Implanted Atrial Defibrillators

The objective of this study is to develop an algorithm able to detect atrial fibrillation episodes monitoring a standard superficial ECG lead. This algorithm aims to be included in a microprocessor based portable monitoring unit able to assess atrial defibrillator performance. The automatic recognition of the atrial fibrillation from superficial ECG is critical mainly because of the small size of the atrial waves as compared to ventricular activity and to their chaotic nature. In this work a morphological detection of the fibrillatory waves to be combined to rhythm information is proposed for a reliable automatic detection of atrial fibrillation/flutter. A pre-processing of the raw ECG aims to cancel the ventricular activity (QRS-T complex). This operation is achieved using an adaptive recurrent filters. Afterwards, the recognition of atrial fibrillation episodes is achieved computing a ratio between predefined bands of the power spectrum. Particular care has been paid in the development of the algorithm to allow real-time operation. The sensitivity and the specificity of the presented algorithm has been tested utilising data provided by the MIT-BIH ECG-database and the IMPROVE data library