Detection of paroxysmal atrial fibrillation: A computationally efficient algorithm for use in a wearable telemedical system

For early detection of the sporadic events of paroxysmal atrial fibrillation (AF), long-term ECG monitoring and data evaluation are required. To allow energy-efficient on-device preprocessing of the data, an as simple as possible algorithm is sought. The proposed approach is based on heart frequency data only and makes use of the increased heart rate variation with the occurrence of AF. Criteria for the choice of the free parameters of this method are derived, and it is shown that, under these conditions, results are insensitive against variations of these parameters. If a compromise between sensitivity and specificity is sought, i. e. equality of these values, 89, 4% are obtained in average. If instead sensitivity alone is optimized, i. e. for use with additional AF detection strategies, an average value of 99.6% is reached with specificity 38.4%. The algorithm forms the first step in the development of a computationally and thus energy efficient signal processing module for the purpose of reducing the amount of data that has to be stored or transmitted and then evaluated by the treating physician.